Fix generation bug: correct positional arg mapping for Quick Generate, main Generate, Save/Load. Add *args unpacking for 80 character inputs.

app.py

@@ -6,7 +6,6 @@ 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
 
@@ -147,24 +146,20 @@ input:focus, textarea:focus, select:focus {
     border-color: #334155 !important;
 }
 
-/* Fix checkbox and label contrast */
 input[type="checkbox"] + label,
 .checkbox-label,
 .gr-checkbox label {
     color: #f8fafc !important;
 }
 
-/* Fix list text in tips */
 li, .prose li, .gr-prose li {
     color: #cbd5e1 !important;
 }
 
-/* Ensure strong/bold text is visible */
 strong, b {
     color: #f8fafc !important;
 }
 
-/* Code inline styling */
 code {
     background: #334155 !important;
     color: #22d3ee !important;
@@ -172,7 +167,6 @@ code {
     border-radius: 4px !important;
 }
 
-/* Progress bar styling */
 progress {
     width: 100%;
     height: 8px;
@@ -188,7 +182,6 @@ progress::-webkit-progress-value {
     border-radius: 4px;
 }
 
-/* Segment list styling */
 .seg-item {
     background: #0f172a;
     border: 1px solid #334155;
@@ -214,10 +207,6 @@ progress::-webkit-progress-value {
 # ---------------------------------------------------------------------------
 
 _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:
@@ -258,15 +247,13 @@ def handle_upload(file_obj) -> tuple:
         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)"
+        chs = pipe.detect_chapters(text)
+        ch_info = " | ".join([f"Ch{c['idx']+1}: {c['word_count']}w" for c in chs[:5]])
+        if len(chs) > 5:
+            ch_info += f" (+{len(chs)-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)'}"
+        return text, f"Loaded {fname} — {wc} words (~{dur}) | {ch_info if chs else '1 section'}"
     except Exception as e:
         return "", f"Error: {e}"
 
@@ -280,11 +267,15 @@ 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)
+def get_chapter_text(text: str, chapter_sel: str) -> str:
+    if not text or chapter_sel == "All" or not chapter_sel:
+        return text
+    try:
+        idx = int(chapter_sel.split(":")[0].replace("Ch", "")) - 1
+        pipe = get_pipeline()
+        return pipe.get_chapter_text(text, idx)
+    except Exception:
+        return text
 
 
 # ---------------------------------------------------------------------------
@@ -295,11 +286,22 @@ def get_chapter_text(text: str, chapter_idx: int) -> str:
 def generate_audiobook_gpu(
     text,
     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,
+    gen_temp, gen_seed, output_fmt, *args
 ):
     if not text or len(text.strip()) < 50:
-        return None, None, "Error: Please provide at least 50 characters of story text.", ""
+        return None, "", "Error: Please provide at least 50 characters of story text.", ""
+
+    # Unpack character args (80 values = 8 chars x 10 fields)
+    names = list(args[0:8])
+    descs = list(args[8:16])
+    modes = list(args[16:24])
+    presets = list(args[24:32])
+    audios = list(args[32:40])
+    ref_texts = list(args[40:48])
+    designs = list(args[48:56])
+    instructs = list(args[56:64])
+    langs = list(args[64:72])
+    speeds = list(args[72:80])
 
     pipe = get_pipeline()
 
@@ -312,7 +314,7 @@ def generate_audiobook_gpu(
         design_desc=nar_design if nar_mode == "design" else None,
         instruct=nar_instruct,
         language=nar_lang,
-        speed=float(nar_speed),
+        speed=float(nar_speed) if nar_speed else 1.0,
     )
 
     char_configs = {}
@@ -348,11 +350,7 @@ def generate_audiobook_gpu(
             temperature=gen_temp,
             seed=int(gen_seed),
         )
-        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"
@@ -361,7 +359,6 @@ def generate_audiobook_gpu(
             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")
@@ -390,7 +387,7 @@ 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),
+        speed=float(speed) if speed else 1.0,
     )
     try:
         wav, sr = pipe.preview_voice(vc)
@@ -425,19 +422,74 @@ def preview_char_voice_gpu(name, mode, preset, audio, ref_text, design, instruct
         return None, f"Preview failed: {e}"
 
 
+# ---------------------------------------------------------------------------
+# Quick Generate
+# ---------------------------------------------------------------------------
+
+@spaces.GPU(duration=180)
+def quick_generate_gpu(text, narrator_preset, gen_temp, output_fmt, gen_seed=42):
+    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)}"
+
+
 # ---------------------------------------------------------------------------
 # 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):
+def do_save_project(text, nar_mode, nar_preset, nar_audio, nar_ref_text, nar_design, nar_instruct, nar_lang, nar_speed, *args):
+    # Unpack character args (80 values) + gen_temp + gen_seed
+    names = list(args[0:8])
+    descs = list(args[8:16])
+    modes = list(args[16:24])
+    presets = list(args[24:32])
+    audios = list(args[32:40])
+    ref_texts = list(args[40:48])
+    designs = list(args[48:56])
+    instructs = list(args[56:64])
+    langs = list(args[64:72])
+    speeds = list(args[72:80])
+    gen_temp = args[80] if len(args) > 80 else 0.7
+    gen_seed = args[81] if len(args) > 81 else 42
+
     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),
+        instruct=nar_instruct, language=nar_lang,
+        speed=float(nar_speed) if nar_speed else 1.0,
     )
     char_configs = {}
     for i in range(8):
@@ -463,7 +515,6 @@ def do_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"),
@@ -475,7 +526,6 @@ def do_load_project(json_str):
             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):
@@ -493,6 +543,8 @@ def do_load_project(json_str):
                     gr.update(value=c.instruct, visible=True),
                     gr.update(value=c.language, visible=True),
                     gr.update(value=c.speed, visible=True),
+                    gr.update(visible=True),
+                    gr.update(visible=True),
                 ])
             else:
                 char_updates.extend([
@@ -507,55 +559,16 @@ def do_load_project(json_str):
                     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)}"
+        return [""] + [gr.update()] * 8 + [gr.update(visible=False)] * 104 + [f"Error loading project: {e}"]
 
 
 # ---------------------------------------------------------------------------
@@ -637,7 +650,7 @@ def build_app():
                         quick_audio = gr.Audio(label="Quick Audiobook", interactive=False)
                         quick_status = gr.Textbox(show_label=False, interactive=False)
                         gr.Markdown("---")
-                        gr.Markdown("**Quick Generate** uses a single narrator voice for the entire text — perfect for articles, essays, and simple stories.")
+                        gr.Markdown("**Quick Generate** uses a single narrator voice for the entire text.")
 
                 with gr.Row():
                     chapter_selector = gr.Dropdown(
@@ -665,7 +678,6 @@ def build_app():
                     outputs=[quick_audio, quick_status],
                 )
 
-                # Chapter detection
                 def refresh_chapters(text):
                     if not text:
                         return gr.update(choices=["All"], value="All")
@@ -703,11 +715,11 @@ def build_app():
 
                     with gr.Column(scale=2):
                         gr.Markdown("## Character Voices")
-                        gr.Markdown("Configure up to 8 characters. Each can use Preset, Clone, or Design mode.")
+                        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_speeds = [], [], [], [], [], []
-                        char_rows, char_preview_btns, char_preview_audios = [], [], []
+                        char_rows, char_preview_btns, char_preview_audios, char_preview_statuses = [], [], [], []
 
                         for i in range(8):
                             visible_default = (i == 0)
@@ -727,12 +739,13 @@ def build_app():
                                 with gr.Row():
                                     cpv_btn = gr.Button("🔊 Preview", variant="secondary", visible=visible_default)
                                     cpv_audio = gr.Audio(label="Preview", interactive=False, visible=visible_default)
+                                    cpv_status = gr.Textbox(show_label=False, 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
+                                    outputs=[cpv_audio, cpv_status],
                                 )
 
                             char_rows.append(row)
@@ -748,6 +761,7 @@ def build_app():
                             char_speeds.append(cspd)
                             char_preview_btns.append(cpv_btn)
                             char_preview_audios.append(cpv_audio)
+                            char_preview_statuses.append(cpv_status)
 
             # ==================== TAB 3: Generate ====================
             with gr.TabItem("⚡ Generate"):
@@ -833,6 +847,7 @@ def build_app():
                         gr.update(value=chars[i].get("speed", 1.0), visible=True),
                         gr.update(visible=True),
                         gr.update(visible=True),
+                        gr.update(visible=True),
                     ])
                 else:
                     updates.extend([
@@ -849,6 +864,7 @@ def build_app():
                         gr.update(visible=False),
                         gr.update(visible=False),
                         gr.update(visible=False),
+                        gr.update(visible=False),
                     ])
             return [status] + updates
 
@@ -856,7 +872,7 @@ def build_app():
             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]]
+                 char_speeds[i], char_preview_btns[i], char_preview_audios[i], char_preview_statuses[i]]
                 for i in range(8)
             ] for item in sublist
         ]
@@ -868,26 +884,16 @@ def build_app():
             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
 
+        def wrapped_generate(story_text, chapter_sel, *args):
+            text = get_chapter_text(story_text, chapter_sel)
+            return generate_audiobook_gpu(text, *args)
+
         gen_btn.click(
             wrapped_generate,
             inputs=gen_inputs,
@@ -895,20 +901,14 @@ def build_app():
         )
 
         # ---------- 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, nar_speed,
-            ] + all_char_inputs + [gen_temp, gen_seed],
-            outputs=[project_json],
-        )
+        save_inputs = [
+            story_input,
+            nar_mode, nar_preset, nar_audio, nar_ref_text, nar_design, nar_instruct, nar_lang, nar_speed,
+        ] + all_char_inputs + [gen_temp, gen_seed]
+        save_btn.click(do_save_project, inputs=save_inputs, 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],
-        )
+        load_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]
+        load_btn.click(do_load_project, inputs=[load_json], outputs=load_outputs)
 
     return demo