Major enhancement: file upload, chapter detection, segment previews, multi-format export, speed control, project save/load, quick generate, sample stories

app.py

@@ -1,9 +1,12 @@
 """
-AudioBook Forge - Gradio Frontend
-High-fidelity audiobook generator with character voice mapping.
+AudioBook Forge - Enhanced Gradio Frontend
+High-fidelity audiobook generator with character voice mapping,
+file upload, chapter selection, segment previews, and project save/load.
 """
 
 import os
+import json
+import base64
 from pathlib import Path
 from typing import Dict, List, Optional
 
@@ -31,6 +34,10 @@ from backend import (
     AudiobookPipeline,
     VoiceConfig,
     PRESET_SPEAKERS,
+    SAMPLE_STORIES,
+    save_project,
+    load_project,
+    estimate_duration,
 )
 
 # ---------------------------------------------------------------------------
@@ -79,6 +86,25 @@ body, .gradio-container {
     padding: 1.25rem !important;
 }
 
+.ab-stat {
+    background: #0f172a;
+    border: 1px solid #334155;
+    border-radius: 10px;
+    padding: 0.75rem 1rem;
+    text-align: center;
+}
+.ab-stat .value {
+    font-size: 1.4rem;
+    font-weight: 700;
+    color: #22d3ee;
+}
+.ab-stat .label {
+    font-size: 0.75rem;
+    color: #94a3b8;
+    text-transform: uppercase;
+    letter-spacing: 0.05em;
+}
+
 button.primary {
     background: linear-gradient(135deg, #6366f1, #4f46e5) !important;
     border: none !important;
@@ -145,6 +171,42 @@ code {
     padding: 0.1rem 0.3rem !important;
     border-radius: 4px !important;
 }
+
+/* Progress bar styling */
+progress {
+    width: 100%;
+    height: 8px;
+    border-radius: 4px;
+    background: #334155;
+}
+progress::-webkit-progress-bar {
+    background: #334155;
+    border-radius: 4px;
+}
+progress::-webkit-progress-value {
+    background: linear-gradient(90deg, #6366f1, #22d3ee);
+    border-radius: 4px;
+}
+
+/* Segment list styling */
+.seg-item {
+    background: #0f172a;
+    border: 1px solid #334155;
+    border-radius: 8px;
+    padding: 0.5rem 0.75rem;
+    margin-bottom: 0.4rem;
+    font-size: 0.85rem;
+}
+.seg-item .seg-type {
+    display: inline-block;
+    padding: 0.1rem 0.4rem;
+    border-radius: 4px;
+    font-size: 0.7rem;
+    font-weight: 600;
+    text-transform: uppercase;
+}
+.seg-type.narration { background: #4f46e5; color: #fff; }
+.seg-type.dialogue { background: #22d3ee; color: #0f172a; }
 """
 
 # ---------------------------------------------------------------------------
@@ -152,6 +214,10 @@ code {
 # ---------------------------------------------------------------------------
 
 _pipeline: Optional[AudiobookPipeline] = None
+_stored_text: str = ""
+_stored_chapters: List[Dict] = []
+_stored_segments_meta: List[Dict] = []
+_stored_segment_paths: List[str] = []
 
 
 def get_pipeline() -> AudiobookPipeline:
@@ -175,6 +241,36 @@ def on_mode_change(mode: str) -> tuple:
         return gr.update(visible=False), gr.update(visible=False), gr.update(visible=False), gr.update(visible=True)
 
 
+def update_stats(text: str) -> tuple:
+    wc = len(text.split()) if text else 0
+    dur = estimate_duration(wc)
+    return str(wc), dur
+
+
+def load_sample(name: str) -> str:
+    return SAMPLE_STORIES.get(name, "")
+
+
+def handle_upload(file_obj) -> tuple:
+    if file_obj is None:
+        return "", "No file uploaded."
+    try:
+        pipe = get_pipeline()
+        text, fname = pipe.parse_upload(file_obj)
+        text = pipe.processor.clean_text(text)
+        global _stored_text, _stored_chapters
+        _stored_text = text
+        _stored_chapters = pipe.detect_chapters(text)
+        ch_info = " | ".join([f"Ch{c['idx']+1}: {c['word_count']}w" for c in _stored_chapters[:5]])
+        if len(_stored_chapters) > 5:
+            ch_info += f" (+{len(_stored_chapters)-5} more)"
+        wc = len(text.split())
+        dur = estimate_duration(wc)
+        return text, f"Loaded {fname} — {wc} words (~{dur}) | Chapters: {ch_info if _stored_chapters else '1 (auto)'}"
+    except Exception as e:
+        return "", f"Error: {e}"
+
+
 def extract_chars(text: str, use_ai: bool) -> tuple:
     if not text or len(text.strip()) < 20:
         return [], "Text too short. Please paste at least a paragraph."
@@ -184,6 +280,13 @@ def extract_chars(text: str, use_ai: bool) -> tuple:
     return chars, status
 
 
+def get_chapter_text(text: str, chapter_idx: int) -> str:
+    if not text:
+        return ""
+    pipe = get_pipeline()
+    return pipe.get_chapter_text(text, chapter_idx)
+
+
 # ---------------------------------------------------------------------------
 # GPU-wrapped functions (ZeroGPU)
 # ---------------------------------------------------------------------------
@@ -191,12 +294,12 @@ def extract_chars(text: str, use_ai: bool) -> tuple:
 @spaces.GPU(duration=180)
 def generate_audiobook_gpu(
     text,
-    nar_mode, nar_preset, nar_audio, nar_ref_text, nar_design, nar_instruct, nar_lang,
-    gen_temp, gen_seed,
-    names, descs, modes, presets, audios, ref_texts, designs, instructs, langs,
+    nar_mode, nar_preset, nar_audio, nar_ref_text, nar_design, nar_instruct, nar_lang, nar_speed,
+    gen_temp, gen_seed, output_fmt,
+    names, descs, modes, presets, audios, ref_texts, designs, instructs, langs, speeds,
 ):
     if not text or len(text.strip()) < 50:
-        return None, "Error: Please provide at least 50 characters of story text."
+        return None, None, "Error: Please provide at least 50 characters of story text.", ""
 
     pipe = get_pipeline()
 
@@ -209,6 +312,7 @@ def generate_audiobook_gpu(
         design_desc=nar_design if nar_mode == "design" else None,
         instruct=nar_instruct,
         language=nar_lang,
+        speed=float(nar_speed),
     )
 
     char_configs = {}
@@ -224,6 +328,7 @@ def generate_audiobook_gpu(
             design_desc=designs[i] if modes[i] == "design" else None,
             instruct=instructs[i] or "",
             language=langs[i],
+            speed=float(speeds[i]) if speeds[i] else 1.0,
         )
         char_configs[names[i]] = vc
 
@@ -235,7 +340,7 @@ def generate_audiobook_gpu(
         print(progress_text)
 
     try:
-        output_path, _ = pipe.generate(
+        output_path, seg_paths, seg_meta = pipe.generate(
             text=text,
             narrator_config=nar_cfg,
             character_configs=char_configs,
@@ -243,15 +348,38 @@ def generate_audiobook_gpu(
             temperature=gen_temp,
             seed=int(gen_seed),
         )
-        return output_path, f"Done! Audiobook generated."
+        global _stored_segment_paths, _stored_segments_meta
+        _stored_segment_paths = seg_paths
+        _stored_segments_meta = seg_meta
+
+        # Build segment list HTML
+        seg_html = "<div style='max-height: 300px; overflow-y: auto;'>"
+        for s in seg_meta[:50]:
+            tclass = "narration" if s['type'] == 'narration' else "dialogue"
+            seg_html += f"<div class='seg-item'><span class='seg-type {tclass}'>{s['type']}</span> <strong>{s['speaker']}</strong>: {s['text']}</div>"
+        if len(seg_meta) > 50:
+            seg_html += f"<div style='text-align:center;color:#94a3b8;padding:0.5rem;'>... and {len(seg_meta)-50} more segments</div>"
+        seg_html += "</div>"
+
+        # Extra export
+        extra_path = None
+        if output_fmt == "wav":
+            extra_path = output_path.replace(".mp3", ".wav")
+            from backend import save_audiobook
+            save_audiobook(seg_paths, extra_path, fmt="wav")
+        elif output_fmt == "zip":
+            extra_path = pipe.export_segments_zip(seg_paths)
+
+        final_path = extra_path if extra_path else output_path
+        return final_path, seg_html, f"Done! {len(seg_meta)} segments generated.", progress_text
     except Exception as e:
         import traceback
         traceback.print_exc()
-        return None, f"Error: {str(e)}"
+        return None, "", f"Error: {str(e)}", progress_text
 
 
 @spaces.GPU(duration=60)
-def preview_narrator_gpu(mode, preset, audio, ref_text, design, instruct, lang):
+def preview_narrator_gpu(mode, preset, audio, ref_text, design, instruct, lang, speed):
     pipe = get_pipeline()
     vc = VoiceConfig(
         name="Narrator",
@@ -262,6 +390,7 @@ def preview_narrator_gpu(mode, preset, audio, ref_text, design, instruct, lang):
         design_desc=design if mode == "design" else None,
         instruct=instruct,
         language=lang,
+        speed=float(speed),
     )
     try:
         wav, sr = pipe.preview_voice(vc)
@@ -272,6 +401,163 @@ def preview_narrator_gpu(mode, preset, audio, ref_text, design, instruct, lang):
         return None, f"Preview failed: {e}"
 
 
+@spaces.GPU(duration=60)
+def preview_char_voice_gpu(name, mode, preset, audio, ref_text, design, instruct, lang, speed):
+    pipe = get_pipeline()
+    vc = VoiceConfig(
+        name=name or "Character",
+        mode=mode,
+        preset=preset if mode == "preset" else None,
+        ref_audio=audio if mode == "clone" and audio else None,
+        ref_text=ref_text if mode == "clone" else None,
+        design_desc=design if mode == "design" else None,
+        instruct=instruct,
+        language=lang,
+        speed=float(speed) if speed else 1.0,
+    )
+    try:
+        sample = f"Hello, I am {name or 'your character'}. This is how I sound in the story."
+        wav, sr = pipe.preview_voice(vc, sample_text=sample)
+        return (sr, wav), f"{name or 'Character'} preview ready!"
+    except Exception as e:
+        import traceback
+        traceback.print_exc()
+        return None, f"Preview failed: {e}"
+
+
+# ---------------------------------------------------------------------------
+# Project Save/Load
+# ---------------------------------------------------------------------------
+
+def do_save_project(text, nar_mode, nar_preset, nar_audio, nar_ref_text, nar_design, nar_instruct, nar_lang, nar_speed,
+                    names, descs, modes, presets, audios, ref_texts, designs, instructs, langs, speeds,
+                    gen_temp, gen_seed):
+    nar_cfg = VoiceConfig(
+        name="Narrator", mode=nar_mode, preset=nar_preset if nar_mode == "preset" else None,
+        ref_audio=nar_audio if nar_mode == "clone" and nar_audio else None,
+        ref_text=nar_ref_text if nar_mode == "clone" else None,
+        design_desc=nar_design if nar_mode == "design" else None,
+        instruct=nar_instruct, language=nar_lang, speed=float(nar_speed),
+    )
+    char_configs = {}
+    for i in range(8):
+        if not names[i]:
+            continue
+        char_configs[names[i]] = VoiceConfig(
+            name=names[i], mode=modes[i],
+            preset=presets[i] if modes[i] == "preset" else None,
+            ref_audio=audios[i] if modes[i] == "clone" and audios[i] else None,
+            ref_text=ref_texts[i] if modes[i] == "clone" else None,
+            design_desc=designs[i] if modes[i] == "design" else None,
+            instruct=instructs[i] or "", language=langs[i],
+            speed=float(speeds[i]) if speeds[i] else 1.0,
+        )
+    settings = {"temperature": gen_temp, "seed": int(gen_seed)}
+    json_str = save_project(text, nar_cfg, char_configs, settings)
+    return json_str
+
+
+def do_load_project(json_str):
+    try:
+        data = load_project(json_str)
+        nar = data["narrator"]
+        chars = data.get("characters", {})
+
+        # Build updates for narrator
+        nar_updates = [
+            gr.update(value=nar.mode),
+            gr.update(value=nar.preset if nar.preset else "Ryan", visible=nar.mode=="preset"),
+            gr.update(value=nar.ref_audio, visible=nar.mode=="clone"),
+            gr.update(value=nar.ref_text, visible=nar.mode=="clone"),
+            gr.update(value=nar.design_desc, visible=nar.mode=="design"),
+            gr.update(value=nar.instruct),
+            gr.update(value=nar.language),
+            gr.update(value=nar.speed),
+        ]
+
+        # Build updates for characters (up to 8)
+        char_updates = []
+        char_items = list(chars.items())[:8]
+        for i in range(8):
+            if i < len(char_items):
+                _, c = char_items[i]
+                char_updates.extend([
+                    gr.update(visible=True),
+                    gr.update(value=c.name, visible=True),
+                    gr.update(value="", visible=True),
+                    gr.update(value=c.mode, visible=True),
+                    gr.update(value=c.preset if c.preset else "Ryan", visible=c.mode=="preset"),
+                    gr.update(value=c.ref_audio, visible=c.mode=="clone"),
+                    gr.update(value=c.ref_text, visible=c.mode=="clone"),
+                    gr.update(value=c.design_desc, visible=c.mode=="design"),
+                    gr.update(value=c.instruct, visible=True),
+                    gr.update(value=c.language, visible=True),
+                    gr.update(value=c.speed, visible=True),
+                ])
+            else:
+                char_updates.extend([
+                    gr.update(visible=False),
+                    gr.update(visible=False),
+                    gr.update(visible=False),
+                    gr.update(visible=False),
+                    gr.update(visible=False),
+                    gr.update(visible=False),
+                    gr.update(visible=False),
+                    gr.update(visible=False),
+                    gr.update(visible=False),
+                    gr.update(visible=False),
+                    gr.update(visible=False),
+                ])
+
+        text_sample = data.get("text_sample", "")
+        return [text_sample] + nar_updates + char_updates + [f"Project loaded! {len(chars)} characters configured."]
+    except Exception as e:
+        return [""] + [gr.update()]*43 + [f"Error loading project: {e}"]
+
+
+# ---------------------------------------------------------------------------
+# Quick Generate
+# ---------------------------------------------------------------------------
+
+@spaces.GPU(duration=180)
+def quick_generate_gpu(text, narrator_preset, gen_temp, gen_seed, output_fmt):
+    """One-click generation with all defaults."""
+    if not text or len(text.strip()) < 50:
+        return None, "Error: Text too short."
+
+    pipe = get_pipeline()
+    nar_cfg = VoiceConfig(name="Narrator", mode="preset", preset=narrator_preset,
+                          language="English", speed=1.0)
+
+    def prog_cb(ratio: float, msg: str):
+        print(f"[{ratio*100:.0f}%] {msg}")
+
+    try:
+        output_path, seg_paths, seg_meta = pipe.generate(
+            text=text,
+            narrator_config=nar_cfg,
+            character_configs={},
+            progress_callback=prog_cb,
+            temperature=gen_temp,
+            seed=int(gen_seed),
+        )
+
+        extra_path = None
+        if output_fmt == "wav":
+            extra_path = output_path.replace(".mp3", ".wav")
+            from backend import save_audiobook
+            save_audiobook(seg_paths, extra_path, fmt="wav")
+        elif output_fmt == "zip":
+            extra_path = pipe.export_segments_zip(seg_paths)
+
+        final_path = extra_path if extra_path else output_path
+        return final_path, f"Quick audiobook ready! {len(seg_meta)} segments."
+    except Exception as e:
+        import traceback
+        traceback.print_exc()
+        return None, f"Error: {str(e)}"
+
+
 # ---------------------------------------------------------------------------
 # Build UI
 # ---------------------------------------------------------------------------
@@ -308,107 +594,126 @@ def build_app():
         """)
 
         with gr.Tabs():
-            # ==================== TAB 1 ====================
-            with gr.TabItem("📖 Story Setup"):
+            # ==================== TAB 1: Story ====================
+            with gr.TabItem("📖 Story"):
                 with gr.Row():
                     with gr.Column(scale=2):
+                        gr.Markdown("### Upload or Paste")
+                        file_upload = gr.File(
+                            label="Upload EPUB, PDF, TXT, or HTML",
+                            file_types=[".txt", ".epub", ".pdf", ".html", ".htm"],
+                        )
                         story_input = gr.TextArea(
                             label="Story Text",
                             placeholder="Paste your book chapter, short story, or script here...",
-                            lines=20,
+                            lines=18,
                             max_lines=40,
                         )
+                        sample_dropdown = gr.Dropdown(
+                            label="Or try a sample story",
+                            choices=list(SAMPLE_STORIES.keys()),
+                            value=None,
+                        )
+
                     with gr.Column(scale=1):
-                        gr.Markdown("### Character Detection")
-                        use_ai_check = gr.Checkbox(
-                            label="Use AI enhancement (slower, more accurate)",
-                            value=False,
+                        gr.Markdown("### Stats")
+                        with gr.Row():
+                            stat_words = gr.Textbox(label="Words", value="0", interactive=False)
+                            stat_dur = gr.Textbox(label="Est. Duration", value="0 sec", interactive=False)
+                        gr.Markdown("---")
+                        gr.Markdown("### Quick Generate")
+                        quick_preset = gr.Dropdown(
+                            choices=list(PRESET_SPEAKERS.keys()),
+                            value="Ryan",
+                            label="Narrator Voice",
+                        )
+                        quick_temp = gr.Slider(minimum=0.1, maximum=1.0, value=0.7, step=0.05, label="Temperature")
+                        quick_fmt = gr.Dropdown(
+                            choices=["mp3", "wav", "zip"],
+                            value="mp3",
+                            label="Output Format",
                         )
-                        extract_btn = gr.Button("🔍 Extract Characters", variant="primary")
+                        quick_btn = gr.Button("⚡ Quick Generate", variant="primary")
+                        quick_audio = gr.Audio(label="Quick Audiobook", interactive=False)
+                        quick_status = gr.Textbox(show_label=False, interactive=False)
                         gr.Markdown("---")
-                        gr.Markdown("**Tips:**")
-                        gr.Markdown("- Use `Character: \"dialogue\"` format for best results.")
-                        gr.Markdown("- Or standard prose with quoted dialogue.")
-                        gr.Markdown("- AI mode uses a small LLM for deeper analysis.")
+                        gr.Markdown("**Quick Generate** uses a single narrator voice for the entire text — perfect for articles, essays, and simple stories.")
+
+                with gr.Row():
+                    chapter_selector = gr.Dropdown(
+                        label="Chapter / Section",
+                        choices=["All"],
+                        value="All",
+                        interactive=True,
+                    )
+                    refresh_chapters_btn = gr.Button("🔄 Detect Chapters")
+
+                with gr.Row():
+                    gr.Markdown("### Character Detection")
+                    use_ai_check = gr.Checkbox(label="Use AI enhancement (slower, more accurate)", value=False)
+                    extract_btn = gr.Button("🔍 Extract Characters", variant="primary")
 
                 extract_status = gr.Textbox(label="Status", interactive=False)
 
-            # ==================== TAB 2 ====================
+                # Wiring
+                file_upload.change(handle_upload, inputs=[file_upload], outputs=[story_input, extract_status])
+                sample_dropdown.change(load_sample, inputs=[sample_dropdown], outputs=[story_input])
+                story_input.change(update_stats, inputs=[story_input], outputs=[stat_words, stat_dur])
+                quick_btn.click(
+                    quick_generate_gpu,
+                    inputs=[story_input, quick_preset, quick_temp, quick_fmt],
+                    outputs=[quick_audio, quick_status],
+                )
+
+                # Chapter detection
+                def refresh_chapters(text):
+                    if not text:
+                        return gr.update(choices=["All"], value="All")
+                    pipe = get_pipeline()
+                    chs = pipe.detect_chapters(text)
+                    choices = ["All"] + [f"Ch{c['idx']+1}: {c['title'][:60]}" for c in chs]
+                    return gr.update(choices=choices, value="All")
+
+                refresh_chapters_btn.click(refresh_chapters, inputs=[story_input], outputs=[chapter_selector])
+
+            # ==================== TAB 2: Voice Cast ====================
             with gr.TabItem("🎭 Voice Cast"):
                 with gr.Row():
                     with gr.Column(scale=1):
                         gr.Markdown("## Narrator")
                         with gr.Column(elem_classes="ab-card"):
-                            nar_mode = gr.Dropdown(
-                                choices=["preset", "clone", "design"],
-                                value="preset",
-                                label="Narrator Mode",
-                            )
-                            nar_preset = gr.Dropdown(
-                                choices=list(PRESET_SPEAKERS.keys()),
-                                value="Ryan",
-                                label="Preset Voice",
-                            )
-                            nar_audio = gr.Audio(
-                                label="Upload Voice Sample (3–10s)",
-                                type="filepath",
-                                visible=False,
-                            )
-                            nar_ref_text = gr.Textbox(
-                                label="Reference Transcript",
-                                placeholder="What does the reference audio say?",
-                                visible=False,
-                            )
-                            nar_design = gr.TextArea(
-                                label="Voice Description",
-                                placeholder="e.g. A warm, raspy baritone with a slight British accent.",
-                                visible=False,
-                                lines=2,
-                            )
-                            nar_instruct = gr.Textbox(
-                                label="Style Instruction",
-                                placeholder="e.g. Calm, measured storytelling pace.",
-                            )
-                            nar_lang = gr.Dropdown(
-                                choices=["English", "Chinese", "Japanese", "Korean", "German", "French", "Spanish", "Italian", "Portuguese", "Russian"],
-                                value="English",
-                                label="Language",
-                            )
+                            nar_mode = gr.Dropdown(choices=["preset", "clone", "design"], value="preset", label="Mode")
+                            nar_preset = gr.Dropdown(choices=list(PRESET_SPEAKERS.keys()), value="Ryan", label="Preset Voice")
+                            nar_audio = gr.Audio(label="Upload Voice Sample (3–10s)", type="filepath", visible=False)
+                            nar_ref_text = gr.Textbox(label="Reference Transcript", placeholder="What does the sample say?", visible=False)
+                            nar_design = gr.TextArea(label="Voice Description", placeholder="e.g. A warm, raspy baritone...", visible=False, lines=2)
+                            nar_instruct = gr.Textbox(label="Style Instruction", placeholder="e.g. Calm, measured storytelling.")
+                            nar_lang = gr.Dropdown(choices=["English", "Chinese", "Japanese", "Korean", "German", "French", "Spanish", "Italian", "Portuguese", "Russian"], value="English", label="Language")
+                            nar_speed = gr.Slider(minimum=0.5, maximum=2.0, value=1.0, step=0.1, label="Speed")
                             nar_preview_btn = gr.Button("🔊 Preview Narrator", variant="secondary")
                             nar_preview_audio = gr.Audio(label="Preview", interactive=False)
                             nar_preview_status = gr.Textbox(show_label=False, interactive=False)
 
-                            nar_mode.change(
-                                on_mode_change,
-                                inputs=nar_mode,
-                                outputs=[nar_preset, nar_audio, nar_ref_text, nar_design],
-                            )
+                            nar_mode.change(on_mode_change, inputs=nar_mode, outputs=[nar_preset, nar_audio, nar_ref_text, nar_design])
                             nar_preview_btn.click(
                                 preview_narrator_gpu,
-                                inputs=[nar_mode, nar_preset, nar_audio, nar_ref_text, nar_design, nar_instruct, nar_lang],
+                                inputs=[nar_mode, nar_preset, nar_audio, nar_ref_text, nar_design, nar_instruct, nar_lang, nar_speed],
                                 outputs=[nar_preview_audio, nar_preview_status],
                             )
 
                     with gr.Column(scale=2):
                         gr.Markdown("## Character Voices")
-                        gr.Markdown("Configure up to 8 characters. Use **preset** for built-in speakers, **clone** to upload a voice sample, or **design** to describe a voice from text.")
-
-                        char_names = []
-                        char_descs = []
-                        char_modes = []
-                        char_presets = []
-                        char_audios = []
-                        char_ref_texts = []
-                        char_designs = []
-                        char_instructs = []
-                        char_langs = []
-                        char_rows = []
+                        gr.Markdown("Configure up to 8 characters. Each can use Preset, Clone, or Design mode.")
+
+                        char_names, char_descs, char_modes, char_presets = [], [], [], []
+                        char_audios, char_ref_texts, char_designs, char_instructs, char_langs, char_speeds = [], [], [], [], [], []
+                        char_rows, char_preview_btns, char_preview_audios = [], [], []
 
                         for i in range(8):
                             visible_default = (i == 0)
                             with gr.Group(visible=visible_default) as row:
                                 with gr.Row():
-                                    cn = gr.Textbox(label=f"Name", placeholder="e.g. Alice", visible=visible_default)
+                                    cn = gr.Textbox(label="Name", placeholder="e.g. Alice", visible=visible_default)
                                     cd = gr.Textbox(label="Description", placeholder="Personality note", visible=visible_default)
                                     cm = gr.Dropdown(label="Mode", choices=["preset", "clone", "design"], value="preset", visible=visible_default)
                                     cp = gr.Dropdown(label="Preset", choices=list(PRESET_SPEAKERS.keys()), value="Ryan", visible=visible_default)
@@ -418,11 +723,16 @@ def build_app():
                                     cdes = gr.TextArea(label="Voice Description", placeholder="e.g. A shrill, nervous teenager.", visible=False, lines=2)
                                     cinstr = gr.Textbox(label="Style Instruction", placeholder="e.g. Angry and loud.", visible=visible_default)
                                     cl = gr.Dropdown(label="Language", choices=["English", "Chinese", "Japanese", "Korean", "German", "French", "Spanish", "Italian", "Portuguese", "Russian"], value="English", visible=visible_default)
-
-                                cm.change(
-                                    on_mode_change,
-                                    inputs=cm,
-                                    outputs=[cp, ca, crt, cdes],
+                                    cspd = gr.Slider(label="Speed", minimum=0.5, maximum=2.0, value=1.0, step=0.1, visible=visible_default)
+                                with gr.Row():
+                                    cpv_btn = gr.Button("🔊 Preview", variant="secondary", visible=visible_default)
+                                    cpv_audio = gr.Audio(label="Preview", interactive=False, visible=visible_default)
+
+                                cm.change(on_mode_change, inputs=cm, outputs=[cp, ca, crt, cdes])
+                                cpv_btn.click(
+                                    preview_char_voice_gpu,
+                                    inputs=[cn, cm, cp, ca, crt, cdes, cinstr, cl, cspd],
+                                    outputs=[cpv_audio, cpv_btn],  # reuse button for status
                                 )
 
                             char_rows.append(row)
@@ -435,46 +745,72 @@ def build_app():
                             char_designs.append(cdes)
                             char_instructs.append(cinstr)
                             char_langs.append(cl)
+                            char_speeds.append(cspd)
+                            char_preview_btns.append(cpv_btn)
+                            char_preview_audios.append(cpv_audio)
 
-            # ==================== TAB 3 ====================
+            # ==================== TAB 3: Generate ====================
             with gr.TabItem("⚡ Generate"):
                 with gr.Row():
                     with gr.Column(scale=1):
                         gr.Markdown("### Settings")
                         gen_temp = gr.Slider(minimum=0.1, maximum=1.0, value=0.7, step=0.05, label="Temperature")
                         gen_seed = gr.Number(value=42, precision=0, label="Seed (fix for consistency)")
-                        gen_btn = gr.Button("▶️ Generate Audiobook", variant="primary", size="lg")
+                        output_fmt = gr.Dropdown(choices=["mp3", "wav", "zip"], value="mp3", label="Output Format")
+                        gen_btn = gr.Button("▶️ Generate Full Audiobook", variant="primary", size="lg")
                         gen_progress = gr.Textbox(label="Progress", interactive=False, value="Ready.")
 
                     with gr.Column(scale=2):
                         gr.Markdown("### Output")
                         output_audio = gr.Audio(label="Generated Audiobook", type="filepath", interactive=False)
                         output_status = gr.Textbox(label="Status", interactive=False)
+                        segment_list = gr.HTML(label="Segments")
 
-            # ==================== TAB 4 ====================
+            # ==================== TAB 4: Project ====================
+            with gr.TabItem("💾 Project"):
+                with gr.Row():
+                    with gr.Column():
+                        gr.Markdown("### Save Project")
+                        save_btn = gr.Button("💾 Save Configuration", variant="primary")
+                        project_json = gr.TextArea(label="Project JSON (copy this to save)", lines=10, interactive=True)
+                    with gr.Column():
+                        gr.Markdown("### Load Project")
+                        load_json = gr.TextArea(label="Paste Project JSON here", lines=10, interactive=True)
+                        load_btn = gr.Button("📂 Load Configuration", variant="secondary")
+                        load_status = gr.Textbox(label="Status", interactive=False)
+
+            # ==================== TAB 5: About ====================
             with gr.TabItem("ℹ️ About"):
                 gr.Markdown("""
                 ## AudioBook Forge
 
-                **Model-agnostic, high-fidelity audiobook generator** powered by [Qwen3-TTS](https://github.com/QwenLM/Qwen3-TTS). Create audiobooks where every character speaks with their own unique voice.
-
-                ## Features
-
-                - 🎙️ **Character Voice Mapping** — Automatically detect characters from your story and assign unique voices to each one
-                - 🎭 **Three Voice Modes**
-                  - **Preset** — 9 premium built-in speakers (English, Chinese, Japanese, Korean, dialects)
-                  - **Clone** — Upload a 3–10 second voice sample to clone any voice
-                  - **Design** — Describe a voice in text and the AI creates it
-                - 📖 **Smart Text Processing** — Automatically distinguishes narration from dialogue and routes each segment to the correct voice
-                - 🌐 **Multilingual** — Supports 10 languages via Qwen3-TTS
-                - ⚡ **ZeroGPU** — Runs on Hugging Face ZeroGPU (free A100 compute)
-                - 🔧 **Model Agnostic** — Backend is swappable; upgrade to future SOTA TTS models without changing the UI
+                **Model-agnostic, high-fidelity audiobook generator** powered by [Qwen3-TTS](https://github.com/QwenLM/Qwen3-TTS).
+
+                ### Features
+                - 📁 **File Upload** — Import EPUB, PDF, TXT, or HTML directly
+                - 📖 **Chapter Detection** — Auto-detects chapters/sections for selective generation
+                - 🎙️ **Character Voice Mapping** — Auto-extract characters and assign unique voices
+                - 🎭 **Three Voice Modes** — Preset (9 speakers), Clone (upload sample), Design (text description)
+                - ⚡ **Quick Generate** — One-click audiobook with a single narrator voice
+                - 🎚️ **Speed Control** — Adjust playback speed per voice (0.5x–2.0x)
+                - 📦 **Multi-format Export** — MP3, WAV, or ZIP of individual segments
+                - 💾 **Save/Load Projects** — Export and restore your voice configurations
+                - 🌐 **10 Languages** — English, Chinese, Japanese, Korean, German, French, Spanish, Italian, Portuguese, Russian
+                - ⚡ **ZeroGPU** — Runs on Hugging Face ZeroGPU (free compute)
+
+                ### Workflow
+                1. **Upload or paste** your story text
+                2. **Detect chapters** (optional) and select a range
+                3. **Extract characters** or use Quick Generate for simple narration
+                4. **Assign voices** to narrator and each character
+                5. **Generate** and download your audiobook
 
                 ### Tips for Best Quality
-                - Use clean, noise-free voice samples for cloning (3–10 seconds).
-                - Keep reference transcripts accurate — they guide the cloning quality.
-                - Lower temperature (0.5–0.6) for stable narration; higher (0.8–0.9) for expressive dialogue.
-                - Use a fixed seed across chunks to prevent voice drift.
+                - Use clean, noise-free voice samples for cloning (3–10 seconds)
+                - Keep reference transcripts accurate
+                - Lower temperature (0.5–0.6) for stable narration; higher (0.8–0.9) for expressive dialogue
+                - Use a fixed seed to prevent voice drift across segments
+                - Use speed adjustment to fine-tune pacing per character
                 """)
 
         # ---------- Extract wiring ----------
@@ -494,6 +830,9 @@ def build_app():
                         gr.update(visible=False),
                         gr.update(value=chars[i].get("voice_instruct", ""), visible=True),
                         gr.update(value=chars[i].get("language", "English"), visible=True),
+                        gr.update(value=chars[i].get("speed", 1.0), visible=True),
+                        gr.update(visible=True),
+                        gr.update(visible=True),
                     ])
                 else:
                     updates.extend([
@@ -507,35 +846,68 @@ def build_app():
                         gr.update(visible=False),
                         gr.update(visible=False),
                         gr.update(visible=False),
+                        gr.update(visible=False),
+                        gr.update(visible=False),
+                        gr.update(visible=False),
                     ])
             return [status] + updates
 
-        extract_btn.click(
-            do_extract,
-            inputs=[story_input, use_ai_check],
-            outputs=[extract_status] + [
-                item for sublist in [
-                    [char_rows[i], char_names[i], char_descs[i], char_modes[i], char_presets[i],
-                     char_audios[i], char_ref_texts[i], char_designs[i], char_instructs[i], char_langs[i]]
-                    for i in range(8)
-                ] for item in sublist
-            ],
-        )
+        extract_outputs = [extract_status] + [
+            item for sublist in [
+                [char_rows[i], char_names[i], char_descs[i], char_modes[i], char_presets[i],
+                 char_audios[i], char_ref_texts[i], char_designs[i], char_instructs[i], char_langs[i],
+                 char_speeds[i], char_preview_btns[i], char_preview_audios[i]]
+                for i in range(8)
+            ] for item in sublist
+        ]
+        extract_btn.click(do_extract, inputs=[story_input, use_ai_check], outputs=extract_outputs)
 
         # ---------- Generate wiring ----------
         all_char_inputs = (
             char_names + char_descs + char_modes + char_presets +
-            char_audios + char_ref_texts + char_designs + char_instructs + char_langs
+            char_audios + char_ref_texts + char_designs + char_instructs + char_langs + char_speeds
         )
 
+        def get_text_for_gen(story_text, chapter_sel):
+            if chapter_sel == "All" or not chapter_sel:
+                return story_text
+            # Extract chapter index
+            try:
+                idx = int(chapter_sel.split(":")[0].replace("Ch", "")) - 1
+                return get_chapter_text(story_text, idx)
+            except:
+                return story_text
+
+        def wrapped_generate(story_text, chapter_sel, *args):
+            text = get_text_for_gen(story_text, chapter_sel)
+            return generate_audiobook_gpu(text, *args)
+
+        gen_inputs = [
+            story_input, chapter_selector,
+            nar_mode, nar_preset, nar_audio, nar_ref_text, nar_design, nar_instruct, nar_lang, nar_speed,
+            gen_temp, gen_seed, output_fmt,
+        ] + all_char_inputs
+
         gen_btn.click(
-            generate_audiobook_gpu,
+            wrapped_generate,
+            inputs=gen_inputs,
+            outputs=[output_audio, segment_list, output_status, gen_progress],
+        )
+
+        # ---------- Project wiring ----------
+        save_btn.click(
+            do_save_project,
             inputs=[
                 story_input,
-                nar_mode, nar_preset, nar_audio, nar_ref_text, nar_design, nar_instruct, nar_lang,
-                gen_temp, gen_seed,
-            ] + all_char_inputs,
-            outputs=[output_audio, output_status],
+                nar_mode, nar_preset, nar_audio, nar_ref_text, nar_design, nar_instruct, nar_lang, nar_speed,
+            ] + all_char_inputs + [gen_temp, gen_seed],
+            outputs=[project_json],
+        )
+
+        load_btn.click(
+            do_load_project,
+            inputs=[load_json],
+            outputs=[story_input, nar_mode, nar_preset, nar_audio, nar_ref_text, nar_design, nar_instruct, nar_lang, nar_speed] + extract_outputs[1:] + [load_status],
         )
 
     return demo

backend.py

@@ -1,7 +1,8 @@
 """
 AudioBook Forge - Backend
 Model-agnostic TTS engine with Qwen3-TTS support.
-Character extraction, dialogue parsing, and audio stitching.
+Character extraction, dialogue parsing, audio stitching, file import,
+chapter detection, segment preview, and multi-format export.
 """
 
 import os
@@ -9,10 +10,12 @@ import re
 import json
 import hashlib
 import tempfile
+import zipfile
 from pathlib import Path
 from typing import List, Dict, Optional, Tuple, Any
-from dataclasses import dataclass, field
+from dataclasses import dataclass, field, asdict
 from collections import defaultdict
+from html.parser import HTMLParser
 import warnings
 
 import numpy as np
@@ -39,6 +42,75 @@ PRESET_SPEAKERS = {
 MAX_CHUNK_CHARS = 380
 MIN_CHUNK_CHARS = 80
 CROSSFADE_MS = 80
+WORDS_PER_MINUTE = 150
+
+SAMPLE_STORIES = {
+    "The Velveteen Rabbit (excerpt)": """There was once a velveteen rabbit, and in the beginning he was really splendid. He was fat and bunchy, as a rabbit should be; his coat was spotted brown and white, he had real thread whiskers, and his ears were lined with pink sateen.
+
+On Christmas morning, when he sat wedged in the top of the Boy's stocking, with a sprig of holly between his paws, the effect was charming.
+
+There were other things in the stocking, nuts and oranges and a toy engine, and chocolate almonds and a clockwork mouse, but the Rabbit was quite the best of all. For at least two hours the Boy loved him, and then Aunts and Uncles came to dinner, and there was a great rustling of tissue paper and unwrapping of parcels, and in the excitement of looking at all the new presents the Velveteen Rabbit was forgotten.
+
+For a long time he lived in the toy cupboard or on the nursery floor, and no one thought very much about him. He was naturally shy, and being only made of velveteen, some of the more expensive toys quite snubbed him. The mechanical toys were very superior, and looked down upon every one else; they were full of modern ideas, and pretended they were real.
+
+The Rabbit could not claim to be a model of anything, for he didn't know that real rabbits existed; he thought they were all stuffed with sawdust like himself, and he understood that sawdust was quite out-of-date and should never be mentioned in modern circles.
+
+Even Timothy, the jointed wooden lion, who was made by the disabled soldiers, and should have had broader views, put on airs and pretended he was connected with Government. Between them all the poor little Rabbit was made to feel himself very insignificant and commonplace, and the only person who was kind to him at all was the Skin Horse.
+
+The Skin Horse had lived longer in the nursery than any of the others. He was so old that his brown coat was bald in patches and showed the seams underneath, and most of the hairs in his tail had been pulled out to string bead necklaces.
+
+He was wise, for he had seen a long succession of mechanical toys arrive to boast and swagger, and by-and-by break their mainsprings and pass away, and he knew that they were only toys, and would never turn into anything else. For nursery magic is very strange and wonderful, and only those playthings that are old and wise and experienced like the Skin Horse understand all about it.""",
+
+    "A Study in Scarlet (excerpt)": """In the year 1878 I took my degree of Doctor of Medicine of the University of London, and proceeded to Netley to go through the course prescribed for surgeons in the army. Having completed my studies there, I was duly attached to the Fifth Northumberland Fusiliers as Assistant Surgeon.
+
+The regiment was stationed in India at the time, and before I could join it, the second Afghan war had broken out. On landing at Bombay, I learned that my corps had advanced through the passes, and was already deep in the enemy's country.
+
+I followed, however, with many other officers who were in the same situation as myself, and succeeded in reaching Candahar in safety, where I found my regiment, and at once entered upon my new duties.
+
+The campaign brought honours and promotion to many, but for me it had nothing but misfortune and disaster. I was removed from my brigade and attached to the Berkshires, with whom I served at the fatal battle of Maiwand.
+
+There I was struck on the shoulder by a Jezail bullet, which shattered the bone and grazed the subclavian artery. I should have fallen into the hands of the murderous Ghazis had it not been for the devotion and courage shown by Murray, my orderly, who threw me across a pack-horse, and succeeded in bringing me safely to the British lines.
+
+Worn with pain, and weak from the prolonged hardships which I had undergone, I was removed, with a great train of wounded sufferers, to the base hospital at Peshawar. Here I rallied, and had already improved so far as to be able to walk about the wards, and even to bask a little upon the verandah, when I was struck down by enteric fever, that curse of our Indian possessions.
+
+For months my life was despaired of, and when at last I came to myself and became convalescent, I was so weak and emaciated that a medical board determined that not a day should be lost in sending me back to England. I was dispatched, accordingly, in the troopship Orontes, and landed a month later on Portsmouth jetty, with my health irretrievably ruined, but with permission from a paternal government to spend the next nine months in attempting to improve it.""",
+
+    "Pride and Prejudice (excerpt)": """It is a truth universally acknowledged, that a single man in possession of a good fortune, must be in want of a wife.
+
+However little known the feelings or views of such a man may be on his first entering a neighbourhood, this truth is so well fixed in the minds of the surrounding families, that he is considered the rightful property of some one or other of their daughters.
+
+\"My dear Mr. Bennet,\" said his lady to him one day, \"have you heard that Netherfield Park is let at last?\"
+
+Mr. Bennet replied that he had not.
+
+\"But it is,\" returned she; \"for Mrs. Long has just been here, and she told me all about it.\"
+
+Mr. Bennet made no answer.
+
+\"Do you not want to know who has taken it?\" cried his wife impatiently.
+
+\"You want to tell me, and I have no objection to hearing it.\"
+
+This was invitation enough.
+
+\"Why, my dear, you must know, Mrs. Long says that Netherfield is taken by a young man of large fortune from the north of England; that he came down on Monday in a chaise and four to see the place, and was so much delighted with it, that he agreed with Mr. Morris immediately; that he is to take possession before Michaelmas, and some of his servants are to be in the house by the end of next week.\"
+
+\"What is his name?\"
+
+\"Bingley.\"
+
+\"Is he married or single?\"
+
+\"Oh! Single, my dear, to be sure! A single man of large fortune; four or five thousand a year. What a fine thing for our girls!\"
+
+\"How so? How can it affect them?\"
+
+\"My dear Mr. Bennet,\" replied his wife, \"how can you be so tiresome! You must know that I am thinking of his marrying one of them.\"
+
+\"Is that his design in settling here?\"
+
+\"Design! Nonsense, how can you talk so! But it is very likely that he may fall in love with one of them, and therefore you must visit him as soon as he comes.\"""",
+}
 
 # ---------------------------------------------------------------------------
 # Data Classes
@@ -47,21 +119,30 @@ CROSSFADE_MS = 80
 @dataclass
 class VoiceConfig:
     name: str = "Narrator"
-    mode: str = "preset"          # preset | clone | design
-    preset: Optional[str] = None   # e.g., "Ryan"
+    mode: str = "preset"
+    preset: Optional[str] = None
     ref_audio: Optional[str] = None
     ref_text: Optional[str] = None
     design_desc: Optional[str] = None
-    instruct: str = ""            # style instruction
+    instruct: str = ""
     language: str = "English"
+    speed: float = 1.0  # 0.5 to 2.0
+
+    def to_dict(self) -> dict:
+        return asdict(self)
+
+    @classmethod
+    def from_dict(cls, d: dict) -> "VoiceConfig":
+        return cls(**{k: v for k, v in d.items() if k in cls.__dataclass_fields__})
 
 
 @dataclass
 class TextSegment:
     text: str
-    seg_type: str                 # narration | dialogue
+    seg_type: str = "narration"
     speaker: Optional[str] = None
     emotion_hint: Optional[str] = None
+    chapter_idx: int = 0
 
 
 @dataclass
@@ -72,16 +153,254 @@ class CharacterProfile:
     occurrences: int = 0
 
 
+@dataclass
+class Chapter:
+    idx: int
+    title: str
+    text: str
+    word_count: int = 0
+
+
 # ---------------------------------------------------------------------------
-# TTS Engine (Model-Agnostic Wrapper)
+# File Importers
 # ---------------------------------------------------------------------------
 
-class TTSEngine:
-    """
-    Model-agnostic TTS engine.
-    Currently backed by Qwen3-TTS. Swappable architecture.
-    """
+class EPUBTextExtractor(HTMLParser):
+    def __init__(self):
+        super().__init__()
+        self.text_parts = []
+        self.in_script = False
+        self.in_body = False
+
+    def handle_starttag(self, tag, attrs):
+        if tag in ("script", "style"):
+            self.in_script = True
+        if tag == "body":
+            self.in_body = True
+        if tag in ("p", "div", "h1", "h2", "h3", "h4", "br"):
+            self.text_parts.append("\n")
+
+    def handle_endtag(self, tag):
+        if tag in ("script", "style"):
+            self.in_script = False
+        if tag in ("p", "div", "h1", "h2", "h3", "h4"):
+            self.text_parts.append("\n")
+
+    def handle_data(self, data):
+        if not self.in_script:
+            self.text_parts.append(data)
+
+    def get_text(self) -> str:
+        text = "".join(self.text_parts)
+        text = re.sub(r"\n\s*\n\s*\n+", "\n\n", text)
+        text = re.sub(r"[ \t]+", " ", text)
+        return text.strip()
+
+
+def parse_file(filepath: str) -> Tuple[str, str]:
+    """Parse uploaded file and return (text, filename)."""
+    path = Path(filepath)
+    suffix = path.suffix.lower()
+
+    if suffix == ".txt":
+        with open(path, "r", encoding="utf-8", errors="ignore") as f:
+            return f.read(), path.name
+
+    elif suffix == ".epub":
+        with zipfile.ZipFile(path, "r") as z:
+            texts = []
+            for name in z.namelist():
+                if name.endswith((".html", ".htm", ".xhtml", ".xml")):
+                    with z.open(name) as f:
+                        content = f.read().decode("utf-8", errors="ignore")
+                        parser = EPUBTextExtractor()
+                        parser.feed(content)
+                        texts.append(parser.get_text())
+            return "\n\n".join(texts), path.name
+
+    elif suffix == ".pdf":
+        try:
+            from PyPDF2 import PdfReader
+            reader = PdfReader(str(path))
+            texts = []
+            for page in reader.pages:
+                t = page.extract_text()
+                if t:
+                    texts.append(t)
+            return "\n\n".join(texts), path.name
+        except Exception as e:
+            raise ValueError(f"PDF parsing failed: {e}")
+
+    elif suffix in (".html", ".htm"):
+        with open(path, "r", encoding="utf-8", errors="ignore") as f:
+            content = f.read()
+            parser = EPUBTextExtractor()
+            parser.feed(content)
+            return parser.get_text(), path.name
+
+    else:
+        raise ValueError(f"Unsupported file type: {suffix}")
+
+
+# ---------------------------------------------------------------------------
+# Text Processing
+# ---------------------------------------------------------------------------
+
+class TextProcessor:
+    DIALOGUE_RE = re.compile(r'(?:^|[.!?\n]\s+)\s*"([^"]{3,500})"')
+    SPEAKER_RE = re.compile(r'(?:^|\n)\s*([A-Z][a-zA-Z\s]{1,20})(?:\s*[:\-–])\s*"([^"]+)"')
+    NAME_RE = re.compile(r'\b([A-Z][a-z]{1,15})\b')
+    CHAPTER_RE = re.compile(
+        r'^(?:\s*(?:Chapter|CHAPTER|Part|PART|Book|BOOK|Section|SECTION)\s*(?:[IVX\d]+|[A-Z]).*)$',
+        re.MULTILINE,
+    )
+    HEADER_RE = re.compile(
+        r'^(?:\s*\d+\s+|\s*Page\s*\d+.*|\s*www\.\S+.*|\s*Copyright.*|\s*All rights reserved.*)$',
+        re.MULTILINE | re.IGNORECASE,
+    )
+
+    @staticmethod
+    def clean_text(text: str) -> str:
+        """Remove headers, page numbers, excessive whitespace."""
+        text = TextProcessor.HEADER_RE.sub("", text)
+        text = re.sub(r"\n\s*\n\s*\n+", "\n\n", text)
+        text = re.sub(r"[ \t]+", " ", text)
+        text = re.sub(r"^\s+", "", text, flags=re.MULTILINE)
+        return text.strip()
+
+    @staticmethod
+    def detect_chapters(text: str) -> List[Chapter]:
+        """Split text into chapters by chapter headings."""
+        matches = list(TextProcessor.CHAPTER_RE.finditer(text))
+        if len(matches) < 2:
+            # No clear chapters; return as single chapter
+            words = len(text.split())
+            return [Chapter(idx=0, title="Full Text", text=text, word_count=words)]
+
+        chapters = []
+        for i, match in enumerate(matches):
+            start = match.start()
+            title = match.group(0).strip()
+            end = matches[i + 1].start() if i + 1 < len(matches) else len(text)
+            ch_text = text[start:end].strip()
+            words = len(ch_text.split())
+            chapters.append(Chapter(idx=i, title=title, text=ch_text, word_count=words))
+        return chapters
+
+    @staticmethod
+    def extract_characters(text: str, use_ai: bool = False) -> List[CharacterProfile]:
+        profiles: Dict[str, CharacterProfile] = {}
+
+        for match in TextProcessor.SPEAKER_RE.finditer(text):
+            name = match.group(1).strip()
+            if len(name) > 2:
+                if name not in profiles:
+                    profiles[name] = CharacterProfile(name=name)
+                profiles[name].occurrences += 1
+
+        for match in TextProcessor.DIALOGUE_RE.finditer(text):
+            before = text[max(0, match.start() - 120):match.start()]
+            said_match = re.search(
+                r'([A-Z][a-z]{1,15})\s+(?:said|cried|shouted|whispered|replied|asked|answered|called|exclaimed)',
+                before,
+            )
+            if said_match:
+                name = said_match.group(1)
+                if name not in profiles:
+                    profiles[name] = CharacterProfile(name=name)
+                profiles[name].occurrences += 1
+
+        all_names = TextProcessor.NAME_RE.findall(text)
+        from collections import Counter
+        common = Counter(all_names).most_common(30)
+        for name, count in common:
+            if count >= 3 and len(name) > 2 and name not in profiles:
+                if name.lower() in {
+                    "the", "and", "but", "for", "are", "was", "were", "had", "have", "has",
+                    "his", "her", "she", "him", "they", "them", "said", "with", "from",
+                    "that", "this", "what", "when", "where", "would", "could", "should",
+                    "not", "you", "all", "any", "can", "had", "her", "was", "one", "our",
+                    "out", "day", "get", "has", "him", "his", "how", "its", "may", "new",
+                    "now", "old", "see", "two", "who", "boy", "man", "way", "too", "upon",
+                }:
+                    continue
+                profiles[name] = CharacterProfile(name=name, occurrences=count)
+
+        result = sorted(profiles.values(), key=lambda p: p.occurrences, reverse=True)
+        return result[:12]
+
+    @staticmethod
+    def segment_text(text: str, characters: List[str]) -> List[TextSegment]:
+        text = text.replace("\r\n", "\n").replace("\r", "\n")
+        paragraphs = [p.strip() for p in re.split(r'\n\s*\n', text) if p.strip()]
+        segments = []
+
+        for para in paragraphs:
+            speaker_match = re.match(
+                r'^([A-Z][a-zA-Z\s]{1,20})[:\-–]\s*"([^"]+)"',
+                para,
+            )
+            if speaker_match:
+                speaker = speaker_match.group(1).strip()
+                dialogue = speaker_match.group(2)
+                segments.append(TextSegment(text=dialogue, seg_type="dialogue", speaker=speaker))
+                remainder = para[speaker_match.end():].strip()
+                if remainder:
+                    segments.append(TextSegment(text=remainder, seg_type="narration"))
+                continue
+
+            parts = re.split(r'"([^"]{3,500})"', para)
+            for i, part in enumerate(parts):
+                part = part.strip()
+                if not part:
+                    continue
+                if i % 2 == 1:
+                    segments.append(TextSegment(text=part, seg_type="dialogue", speaker=None))
+                else:
+                    segments.append(TextSegment(text=part, seg_type="narration"))
+
+        merged = []
+        for seg in segments:
+            if merged and seg.seg_type == "narration" and merged[-1].seg_type == "narration":
+                merged[-1].text += " " + seg.text
+            else:
+                merged.append(seg)
+        return merged
+
+    @staticmethod
+    def chunk_segments(segments: List[TextSegment], max_chars: int = MAX_CHUNK_CHARS) -> List[TextSegment]:
+        result = []
+        for seg in segments:
+            if len(seg.text) <= max_chars:
+                result.append(seg)
+                continue
+            sentences = re.split(r'(?<=[.!?])\s+', seg.text)
+            current_text = ""
+            for sent in sentences:
+                if len(current_text) + len(sent) + 1 <= max_chars:
+                    current_text += (" " if current_text else "") + sent
+                else:
+                    if current_text:
+                        result.append(TextSegment(
+                            text=current_text.strip(),
+                            seg_type=seg.seg_type,
+                            speaker=seg.speaker,
+                        ))
+                    current_text = sent
+            if current_text:
+                result.append(TextSegment(
+                    text=current_text.strip(),
+                    seg_type=seg.seg_type,
+                    speaker=seg.speaker,
+                ))
+        return result
+
+
+# ---------------------------------------------------------------------------
+# TTS Engine
+# ---------------------------------------------------------------------------
 
+class TTSEngine:
     def __init__(self, device: str = "cuda"):
         self.device = device
         self._custom_voice_model = None
@@ -150,7 +469,10 @@ class TTSEngine:
         return self._design_model
 
     def _cache_key(self, text: str, voice: VoiceConfig) -> str:
-        payload = f"{text}|{voice.mode}|{voice.preset}|{voice.ref_audio}|{voice.design_desc}|{voice.instruct}|{voice.language}"
+        payload = (
+            f"{text}|{voice.mode}|{voice.preset}|{voice.ref_audio}|"
+            f"{voice.design_desc}|{voice.instruct}|{voice.language}|{voice.speed}"
+        )
         return hashlib.md5(payload.encode()).hexdigest()
 
     def _cached_path(self, key: str) -> Path:
@@ -163,7 +485,6 @@ class TTSEngine:
         temperature: float = 0.7,
         seed: int = 42,
     ) -> Tuple[np.ndarray, int]:
-        """Generate audio for a text chunk. Returns (audio_array, sample_rate)."""
         cache_key = self._cache_key(text, voice)
         cache_path = self._cached_path(cache_key)
         if cache_path.exists():
@@ -205,15 +526,43 @@ class TTSEngine:
         else:
             raise ValueError(f"Unknown voice mode: {voice.mode}")
 
-        # Handle stereo or list returns
         if isinstance(wavs, list):
             wavs = wavs[0]
         if wavs.ndim > 1:
             wavs = wavs.mean(axis=1)
 
+        # Apply speed adjustment
+        if voice.speed != 1.0 and voice.speed > 0.3:
+            wavs = self._adjust_speed(wavs, sr, voice.speed)
+
         sf.write(str(cache_path), wavs, sr)
         return wavs, sr
 
+    @staticmethod
+    def _adjust_speed(audio: np.ndarray, sr: int, speed: float) -> np.ndarray:
+        """Adjust audio speed using pydub."""
+        if abs(speed - 1.0) < 0.05:
+            return audio
+        # Convert to pydub AudioSegment
+        audio = (audio * 32767).astype(np.int16)
+        seg = AudioSegment(
+            audio.tobytes(),
+            frame_rate=sr,
+            sample_width=2,
+            channels=1 if audio.ndim == 1 else audio.shape[1],
+        )
+        if speed > 1.0:
+            seg = seg.speedup(playback_speed=speed)
+        else:
+            # slowdown by adding frames
+            seg = seg._spawn(seg.raw_data, overrides={
+                "frame_rate": int(seg.frame_rate * speed)
+            })
+            seg = seg.set_frame_rate(sr)
+        # Convert back to numpy
+        samples = np.array(seg.get_array_of_samples())
+        return samples.astype(np.float32) / 32767.0
+
     def status(self) -> Dict[str, Any]:
         return {
             "custom_loaded": self._custom_voice_model is not None,
@@ -222,144 +571,16 @@ class TTSEngine:
         }
 
 
-# ---------------------------------------------------------------------------
-# Text Processing
-# ---------------------------------------------------------------------------
-
-class TextProcessor:
-    """Extract characters, parse dialogue, chunk text."""
-
-    DIALOGUE_RE = re.compile(
-        r'(?:^|[.!?\n]\s+)\s*"([^"]{3,500})"'  # quoted dialogue
-    )
-    SPEAKER_RE = re.compile(
-        r'(?:^|\n)\s*([A-Z][a-zA-Z\s]{1,20})(?:\s*[:\-–])\s*"([^"]+)"'
-    )
-    NAME_RE = re.compile(
-        r'\b([A-Z][a-z]{1,15})\b'
-    )
-
-    @staticmethod
-    def extract_characters(text: str, use_ai: bool = False) -> List[CharacterProfile]:
-        """Extract character names and basic stats from text."""
-        profiles: Dict[str, CharacterProfile] = {}
-
-        # Pattern: Name: "dialogue"
-        for match in TextProcessor.SPEAKER_RE.finditer(text):
-            name = match.group(1).strip()
-            if len(name) > 2:
-                if name not in profiles:
-                    profiles[name] = CharacterProfile(name=name)
-                profiles[name].occurrences += 1
-
-        # Pattern: quoted dialogue near "he said / she said"
-        for match in TextProcessor.DIALOGUE_RE.finditer(text):
-            quote = match.group(1)
-            before = text[max(0, match.start() - 120):match.start()]
-            said_match = re.search(r'([A-Z][a-z]{1,15})\s+(?:said|cried|shouted|whispered|replied|asked)', before)
-            if said_match:
-                name = said_match.group(1)
-                if name not in profiles:
-                    profiles[name] = CharacterProfile(name=name)
-                profiles[name].occurrences += 1
-
-        # Fallback: capitalized names appearing frequently
-        all_names = TextProcessor.NAME_RE.findall(text)
-        from collections import Counter
-        common = Counter(all_names).most_common(30)
-        for name, count in common:
-            if count >= 3 and len(name) > 2 and name not in profiles:
-                # Filter common words
-                if name.lower() in {"the", "and", "but", "for", "are", "was", "were", "had", "have", "has", "his", "her", "she", "him", "they", "them", "said", "with", "from", "that", "this", "what", "when", "where", "would", "could", "should"}:
-                    continue
-                profiles[name] = CharacterProfile(name=name, occurrences=count)
-
-        result = sorted(profiles.values(), key=lambda p: p.occurrences, reverse=True)
-        return result[:12]  # Cap at 12 characters
-
-    @staticmethod
-    def segment_text(text: str, characters: List[str]) -> List[TextSegment]:
-        """Split text into narration/dialogue segments."""
-        segments = []
-        # Normalize newlines
-        text = text.replace("\r\n", "\n").replace("\r", "\n")
-
-        # Split by paragraphs first
-        paragraphs = [p.strip() for p in re.split(r'\n\s*\n', text) if p.strip()]
-
-        for para in paragraphs:
-            # Check if paragraph starts with Character: "dialogue"
-            speaker_match = re.match(r'^([A-Z][a-zA-Z\s]{1,20})[:\-–]\s*"([^"]+)"', para)
-            if speaker_match:
-                speaker = speaker_match.group(1).strip()
-                dialogue = speaker_match.group(2)
-                segments.append(TextSegment(text=dialogue, seg_type="dialogue", speaker=speaker))
-                # Remainder of paragraph as narration
-                remainder = para[speaker_match.end():].strip()
-                if remainder:
-                    segments.append(TextSegment(text=remainder, seg_type="narration"))
-                continue
-
-            # Check for inline quotes
-            parts = re.split(r'"([^"]{3,500})"', para)
-            for i, part in enumerate(parts):
-                part = part.strip()
-                if not part:
-                    continue
-                if i % 2 == 1:
-                    # This was inside quotes
-                    # Try to attribute speaker from surrounding text
-                    speaker = None
-                    segments.append(TextSegment(text=part, seg_type="dialogue", speaker=speaker))
-                else:
-                    segments.append(TextSegment(text=part, seg_type="narration"))
-
-        # Merge adjacent narration segments
-        merged = []
-        for seg in segments:
-            if merged and seg.seg_type == "narration" and merged[-1].seg_type == "narration":
-                merged[-1].text += " " + seg.text
-            else:
-                merged.append(seg)
-        return merged
-
-    @staticmethod
-    def chunk_segments(segments: List[TextSegment], max_chars: int = MAX_CHUNK_CHARS) -> List[TextSegment]:
-        """Break long segments into smaller chunks at sentence boundaries."""
-        result = []
-        for seg in segments:
-            if len(seg.text) <= max_chars:
-                result.append(seg)
-                continue
-            # Split into sentences
-            sentences = re.split(r'(?<=[.!?])\s+', seg.text)
-            current_text = ""
-            current_speaker = seg.speaker
-            current_type = seg.seg_type
-            for sent in sentences:
-                if len(current_text) + len(sent) + 1 <= max_chars:
-                    current_text += (" " if current_text else "") + sent
-                else:
-                    if current_text:
-                        result.append(TextSegment(text=current_text.strip(), seg_type=current_type, speaker=current_speaker))
-                    current_text = sent
-            if current_text:
-                result.append(TextSegment(text=current_text.strip(), seg_type=current_type, speaker=current_speaker))
-        return result
-
-
 # ---------------------------------------------------------------------------
 # Audio Utils
 # ---------------------------------------------------------------------------
 
 def stitch_audio(paths: List[str], crossfade_ms: int = CROSSFADE_MS) -> AudioSegment:
-    """Concatenate WAV files with crossfade."""
     if not paths:
         return AudioSegment.silent(duration=0)
     combined = AudioSegment.from_wav(paths[0])
     for p in paths[1:]:
         next_seg = AudioSegment.from_wav(p)
-        # Simple overlap crossfade
         if crossfade_ms > 0 and len(combined) > crossfade_ms and len(next_seg) > crossfade_ms:
             combined = combined.append(next_seg, crossfade=crossfade_ms)
         else:
@@ -368,41 +589,91 @@ def stitch_audio(paths: List[str], crossfade_ms: int = CROSSFADE_MS) -> AudioSeg
 
 
 def normalize_audio(audio: AudioSegment, target_dBFS: float = -1.5) -> AudioSegment:
-    """Peak normalize audio."""
     change = target_dBFS - audio.max_dBFS
     return audio.apply_gain(change)
 
 
-def save_audiobook(segments_paths: List[str], output_path: str, title: str = "Audiobook") -> str:
-    """Stitch segments and export final audiobook."""
+def save_audiobook(segments_paths: List[str], output_path: str, title: str = "Audiobook", fmt: str = "mp3") -> str:
     if not segments_paths:
         return ""
     combined = stitch_audio(segments_paths)
     combined = normalize_audio(combined)
-    combined.export(output_path, format="mp3", bitrate="192k", tags={"title": title, "artist": "AudioBook Forge"})
+    if fmt == "mp3":
+        combined.export(output_path, format="mp3", bitrate="192k", tags={"title": title, "artist": "AudioBook Forge"})
+    elif fmt == "wav":
+        combined.export(output_path, format="wav", tags={"title": title, "artist": "AudioBook Forge"})
+    elif fmt == "ogg":
+        combined.export(output_path, format="ogg", tags={"title": title, "artist": "AudioBook Forge"})
+    return output_path
+
+
+def save_segment_zip(segments_paths: List[str], output_path: str) -> str:
+    """Save individual segment WAVs as a ZIP."""
+    with zipfile.ZipFile(output_path, "w", zipfile.ZIP_DEFLATED) as zf:
+        for i, p in enumerate(segments_paths):
+            arcname = f"segment_{i:04d}.wav"
+            zf.write(p, arcname)
     return output_path
 
 
+def estimate_duration(word_count: int, wpm: int = WORDS_PER_MINUTE) -> str:
+    minutes = word_count / wpm
+    if minutes < 1:
+        return f"{int(minutes * 60)} seconds"
+    elif minutes < 60:
+        return f"{minutes:.1f} minutes"
+    else:
+        hours = int(minutes // 60)
+        mins = int(minutes % 60)
+        return f"{hours}h {mins}m"
+
+
+# ---------------------------------------------------------------------------
+# Project Save/Load
+# ---------------------------------------------------------------------------
+
+def save_project(
+    text: str,
+    narrator: VoiceConfig,
+    characters: Dict[str, VoiceConfig],
+    settings: dict,
+) -> str:
+    """Save project to JSON string."""
+    data = {
+        "version": "1.1",
+        "text_sample": text[:2000] + ("..." if len(text) > 2000 else ""),
+        "text_hash": hashlib.md5(text.encode()).hexdigest()[:16],
+        "narrator": narrator.to_dict(),
+        "characters": {k: v.to_dict() for k, v in characters.items()},
+        "settings": settings,
+    }
+    return json.dumps(data, indent=2)
+
+
+def load_project(json_str: str) -> dict:
+    """Load project from JSON string."""
+    data = json.loads(json_str)
+    if data.get("version", "1.0").startswith("1."):
+        data["narrator"] = VoiceConfig.from_dict(data["narrator"])
+        data["characters"] = {k: VoiceConfig.from_dict(v) for k, v in data.get("characters", {}).items()}
+    return data
+
+
 # ---------------------------------------------------------------------------
-# Optional: AI Character Extraction via HF Inference
+# AI Character Extraction
 # ---------------------------------------------------------------------------
 
 def ai_extract_characters(text: str, api_token: Optional[str] = None) -> List[CharacterProfile]:
-    """Use a small HF model to extract characters with descriptions."""
     try:
         from huggingface_hub import InferenceClient
         client = InferenceClient(token=api_token or os.getenv("HF_TOKEN"))
-
-        # Truncate text for context window
         sample = text[:4000] + ("\n...[truncated]" if len(text) > 4000 else "")
-
         prompt = (
             "Extract all named characters from the following story excerpt. "
             "For each character, provide their name and a brief description of their personality/role. "
             "Return ONLY a JSON array like: [{\"name\":\"Alice\",\"description\":\"Curious young girl\"},...]\n\n"
             f"STORY:\n{sample}\n\nJSON:"
         )
-
         response = client.text_generation(
             model="Qwen/Qwen3-1.7B",
             prompt=prompt,
@@ -410,8 +681,6 @@ def ai_extract_characters(text: str, api_token: Optional[str] = None) -> List[Ch
             temperature=0.3,
             return_full_text=False,
         )
-
-        # Extract JSON from response
         json_match = re.search(r'\[.*?\]', response, re.DOTALL)
         if json_match:
             data = json.loads(json_match.group())
@@ -438,6 +707,22 @@ class AudiobookPipeline:
         self.temp_dir = Path(tempfile.gettempdir()) / "audiobook_segments"
         self.temp_dir.mkdir(exist_ok=True)
 
+    def parse_upload(self, filepath: str) -> Tuple[str, str]:
+        return parse_file(filepath)
+
+    def detect_chapters(self, text: str) -> List[Dict]:
+        chapters = self.processor.detect_chapters(text)
+        return [
+            {"idx": c.idx, "title": c.title, "word_count": c.word_count}
+            for c in chapters
+        ]
+
+    def get_chapter_text(self, text: str, chapter_idx: int) -> str:
+        chapters = self.processor.detect_chapters(text)
+        if 0 <= chapter_idx < len(chapters):
+            return chapters[chapter_idx].text
+        return text
+
     def extract_characters(self, text: str, use_ai: bool = False) -> List[Dict]:
         if use_ai:
             profiles = ai_extract_characters(text)
@@ -453,10 +738,21 @@ class AudiobookPipeline:
                 "voice_mode": "preset",
                 "voice_preset": "Ryan",
                 "voice_instruct": "",
+                "speed": 1.0,
+                "language": "English",
             }
             for p in profiles
         ]
 
+    def preview_segment(
+        self,
+        text: str,
+        voice: VoiceConfig,
+        temperature: float = 0.7,
+        seed: int = 42,
+    ) -> Tuple[np.ndarray, int]:
+        return self.tts.synthesize(text, voice, temperature=temperature, seed=seed)
+
     def generate(
         self,
         text: str,
@@ -465,22 +761,22 @@ class AudiobookPipeline:
         progress_callback=None,
         temperature: float = 0.7,
         seed: int = 42,
-    ) -> Tuple[str, List[str]]:
+    ) -> Tuple[str, List[str], List[Dict]]:
         """
         Generate audiobook.
-        Returns (final_mp3_path, list_of_segment_wav_paths).
+        Returns (final_path, segment_paths, segment_metadata).
         """
         segments = self.processor.segment_text(text, list(character_configs.keys()))
         segments = self.processor.chunk_segments(segments)
 
         segment_paths = []
+        segment_meta = []
         total = len(segments)
 
         for i, seg in enumerate(segments):
             if progress_callback:
-                progress_callback(i / total, f"Generating segment {i+1}/{total} ({seg.seg_type})...")
+                progress_callback(i / total, f"Segment {i+1}/{total} ({seg.seg_type})...")
 
-            # Determine voice
             if seg.seg_type == "dialogue" and seg.speaker and seg.speaker in character_configs:
                 voice = character_configs[seg.speaker]
             else:
@@ -491,20 +787,38 @@ class AudiobookPipeline:
                 seg_path = self.temp_dir / f"seg_{i:04d}_{voice.name}.wav"
                 sf.write(str(seg_path), wav, sr)
                 segment_paths.append(str(seg_path))
+                segment_meta.append({
+                    "idx": i,
+                    "type": seg.seg_type,
+                    "speaker": seg.speaker or voice.name,
+                    "text": seg.text[:100] + ("..." if len(seg.text) > 100 else ""),
+                    "path": str(seg_path),
+                })
             except Exception as e:
                 print(f"[Pipeline] Segment {i} failed: {e}")
-                # Insert silence to maintain timing
                 silent = AudioSegment.silent(duration=500)
                 seg_path = self.temp_dir / f"seg_{i:04d}_silent.wav"
                 silent.export(str(seg_path), format="wav")
                 segment_paths.append(str(seg_path))
+                segment_meta.append({
+                    "idx": i,
+                    "type": seg.seg_type,
+                    "speaker": voice.name,
+                    "text": seg.text[:100] + "...",
+                    "path": str(seg_path),
+                    "error": str(e),
+                })
 
         if progress_callback:
-            progress_callback(1.0, "Stitching final audiobook...")
+            progress_callback(1.0, "Finalizing audiobook...")
 
         output_path = str(self.temp_dir / "audiobook_final.mp3")
         save_audiobook(segment_paths, output_path, title="Generated Audiobook")
-        return output_path, segment_paths
+        return output_path, segment_paths, segment_meta
+
+    def export_segments_zip(self, segment_paths: List[str]) -> str:
+        output_path = str(self.temp_dir / "audiobook_segments.zip")
+        return save_segment_zip(segment_paths, output_path)
 
     def preview_voice(
         self,

requirements.txt

@@ -9,3 +9,4 @@ huggingface-hub>=0.23.0
 soundfile>=0.12.0
 pydub>=0.25.0
 numpy>=1.26.0
+PyPDF2>=3.0.0