Upload app.py

app.py

@@ -110,17 +110,46 @@ def generate_silence(dur, path):
                    capture_output=True, check=True)
 
 
+# ==========================================
+# TEXT RESOLVER (shared across all modes)
+# ==========================================
+def resolve_text(text_input, file_input):
+    """Get text from either the textbox or an uploaded file."""
+    if file_input is not None:
+        ext = os.path.splitext(file_input)[1].lower()
+        if ext == ".pdf":
+            try:
+                import pypdf
+                reader = pypdf.PdfReader(file_input)
+                return "\n\n".join(p.extract_text().strip() for p in reader.pages if p.extract_text())
+            except Exception as e:
+                raise gr.Error(f"PDF read failed: {e}")
+        elif ext == ".docx":
+            try:
+                import docx
+                doc = docx.Document(file_input)
+                return "\n\n".join(p.text.strip() for p in doc.paragraphs if p.text.strip())
+            except Exception as e:
+                raise gr.Error(f"DOCX read failed: {e}")
+        else:
+            with open(file_input, "r", encoding="utf-8", errors="replace") as f:
+                return f.read()
+    elif text_input and text_input.strip():
+        return text_input.strip()
+    else:
+        raise gr.Error("Please enter text or upload a file.")
+
+
 # ==========================================
 # SINGLE-SPEAKER MODES
 # ==========================================
 @spaces.GPU
-def generate_custom_voice(text, language, speaker_label, instruction):
-    if not text.strip():
-        raise gr.Error("Please enter some text.")
+def generate_custom_voice(text, file_input, language, speaker_label, instruction):
+    resolved_text = resolve_text(text, file_input)
     model = get_model("custom")
     speaker = speaker_label.split("--")[0].strip()
     lang = language if language != "Auto" else "Auto"
-    kwargs = {"text": text, "language": lang, "speaker": speaker}
+    kwargs = {"text": resolved_text, "language": lang, "speaker": speaker}
     if instruction and instruction.strip():
         kwargs["instruct"] = instruction.strip()
     print(f"[TTS] Custom: speaker={speaker}, lang={lang}")
@@ -131,15 +160,14 @@ def generate_custom_voice(text, language, speaker_label, instruction):
 
 
 @spaces.GPU
-def generate_voice_design(text, language, voice_description):
-    if not text.strip():
-        raise gr.Error("Please enter some text.")
+def generate_voice_design(text, file_input, language, voice_description):
+    resolved_text = resolve_text(text, file_input)
     if not voice_description.strip():
         raise gr.Error("Please describe the voice you want.")
     model = get_model("design")
     lang = language if language != "Auto" else "Auto"
     print(f"[TTS] Design: lang={lang}, desc={voice_description[:60]}")
-    wavs, sr = model.generate_voice_design(text=text, language=lang, instruct=voice_description)
+    wavs, sr = model.generate_voice_design(text=resolved_text, language=lang, instruct=voice_description)
     path = os.path.join(OUTPUT_DIR, f"design_{int(time.time())}.wav")
     sf.write(path, wavs[0], sr)
     return path
@@ -244,14 +272,13 @@ def extract_pdf_sections(filepath):
 
 
 @spaces.GPU
-def generate_voice_clone(text, language, ref_audio, ref_text, add_emotions):
-    if not text.strip():
-        raise gr.Error("Please enter some text.")
+def generate_voice_clone(text, file_input, language, ref_audio, ref_text, add_emotions):
+    resolved_text = resolve_text(text, file_input)
     if ref_audio is None:
         raise gr.Error("Please upload a reference audio sample.")
 
     # Enhance text with emotions if requested
-    final_text = text
+    final_text = resolved_text
     if add_emotions:
         client = get_llm_client()
         if client:
@@ -574,7 +601,9 @@ with gr.Blocks(title="Qwen3-TTS Demo") as demo:
         gr.Markdown("Pick a preset speaker with optional emotion/style.")
         with gr.Row():
             with gr.Column():
-                cv_text = gr.Textbox(label="Text", lines=4, placeholder="Enter text...")
+                cv_text = gr.Textbox(label="Text", lines=4, placeholder="Enter text or upload a file below...")
+                cv_file = gr.File(label="Or Upload (.txt, .pdf, .docx)",
+                                  file_types=[".txt", ".md", ".pdf", ".docx"], type="filepath")
                 cv_lang = gr.Dropdown(choices=LANGUAGES, value="Auto", label="Language")
                 cv_speaker = gr.Dropdown(choices=SPEAKER_CHOICES,
                                          value="Ryan -- Dynamic male, strong rhythmic drive (English)",
@@ -585,14 +614,16 @@ with gr.Blocks(title="Qwen3-TTS Demo") as demo:
             with gr.Column():
                 cv_audio = gr.Audio(label="Generated Speech", type="filepath")
         cv_btn.click(fn=generate_custom_voice,
-                     inputs=[cv_text, cv_lang, cv_speaker, cv_instruct], outputs=cv_audio)
+                     inputs=[cv_text, cv_file, cv_lang, cv_speaker, cv_instruct], outputs=cv_audio)
 
     # ── Tab 2: Voice Design ──
     with gr.Tab("Voice Design"):
         gr.Markdown("Describe the voice you want — the AI creates it.")
         with gr.Row():
             with gr.Column():
-                vd_text = gr.Textbox(label="Text", lines=4, placeholder="Enter text...")
+                vd_text = gr.Textbox(label="Text", lines=4, placeholder="Enter text or upload a file below...")
+                vd_file = gr.File(label="Or Upload (.txt, .pdf, .docx)",
+                                  file_types=[".txt", ".md", ".pdf", ".docx"], type="filepath")
                 vd_lang = gr.Dropdown(choices=LANGUAGES, value="English", label="Language")
                 vd_desc = gr.Textbox(label="Voice Description", lines=3,
                                      placeholder="e.g. Warm male storyteller with British accent...")
@@ -610,7 +641,7 @@ with gr.Blocks(title="Qwen3-TTS Demo") as demo:
             with gr.Column():
                 vd_audio = gr.Audio(label="Generated Speech", type="filepath")
         vd_btn.click(fn=generate_voice_design,
-                     inputs=[vd_text, vd_lang, vd_desc], outputs=vd_audio)
+                     inputs=[vd_text, vd_file, vd_lang, vd_desc], outputs=vd_audio)
 
     # ── Tab 3: Voice Clone ──
     with gr.Tab("Voice Clone"):
@@ -620,18 +651,20 @@ with gr.Blocks(title="Qwen3-TTS Demo") as demo:
         )
         with gr.Row():
             with gr.Column():
-                vc_text = gr.Textbox(label="Text (in cloned voice)", lines=4, placeholder="What should the cloned voice say...")
+                vc_text = gr.Textbox(label="Text (in cloned voice)", lines=4, placeholder="Enter text or upload a file below...")
+                vc_file = gr.File(label="Or Upload (.txt, .pdf, .docx)",
+                                  file_types=[".txt", ".md", ".pdf", ".docx"], type="filepath")
                 vc_lang = gr.Dropdown(choices=LANGUAGES, value="English", label="Language")
                 vc_ref = gr.Audio(label="Reference Audio (3+ seconds)", type="filepath")
                 vc_ref_text = gr.Textbox(label="Transcript (optional, improves quality)",
                                          placeholder="What the person says in the reference...")
                 vc_emotions = gr.Checkbox(value=False, label="Enhance with emotions",
-                                          info="AI adds natural emotional cues (pauses, emphasis, interjections) to make the delivery more expressive. Requires DASHSCOPE_API_KEY.")
+                                          info="AI adds natural emotional cues. Requires DASHSCOPE_API_KEY.")
                 vc_btn = gr.Button("Clone & Generate", variant="primary")
             with gr.Column():
                 vc_audio = gr.Audio(label="Cloned Voice Speech", type="filepath")
         vc_btn.click(fn=generate_voice_clone,
-                     inputs=[vc_text, vc_lang, vc_ref, vc_ref_text, vc_emotions], outputs=vc_audio)
+                     inputs=[vc_text, vc_file, vc_lang, vc_ref, vc_ref_text, vc_emotions], outputs=vc_audio)
 
     # ── Tab 4: Multi-Speaker Story ──
     with gr.Tab("Multi-Speaker Story"):