webui/app.py

@@ -252,6 +252,7 @@ def apply_edit(
     current_ref: str,
     history: list,
     enhance: str,
+    timbre_description: str,
     emotion: str,
     other_emotion: str,
     pitch: str,
@@ -266,6 +267,14 @@ def apply_edit(
     # character/accent editing, not after (garbage in, garbage timbre-edited).
     if enhance and ENHANCE_INSTRUCTIONS.get(enhance):
         instructions.append(ENHANCE_INSTRUCTIONS[enhance])
+    if timbre_description and timbre_description.strip():
+        # This is AuK's Timbre task - the one actually used to build every
+        # character's base voice in this project (accent, age, register,
+        # personality all described in one free-form sentence). It has no
+        # fixed vocabulary, unlike Emotion below.
+        instructions.append(
+            f'Keep the spoken content unchanged and change the timbre to: "{timbre_description.strip()}".'
+        )
     if emotion and emotion != "none":
         instructions.append(auk_emotion_instruction(emotion))
     if other_emotion and other_emotion.strip():
@@ -577,13 +586,26 @@ with gr.Blocks(title="Audio-demo pipeline studio") as demo:
                         list(ENHANCE_INSTRUCTIONS.keys()), value="none",
                         label="Enhance (denoise/dereverb/restore) - runs first, before character edits",
                     )
+                    timbre_tb = gr.Textbox(
+                        label="Timbre / voice description (accent, age, register, personality)",
+                        placeholder=(
+                            'e.g. "British male, mid-30s, dry and understated, slight Essex '
+                            'accent, deadpan delivery" - this is what actually built every '
+                            "character's base voice in this project. Free-form, no fixed "
+                            "vocabulary (unlike Emotion below)."
+                        ),
+                        lines=2,
+                    )
                     emotion_dd = gr.Dropdown(
                         ["none"] + EMOTIONS, value="none",
                         label="Emotion (AuK's fixed vocabulary - anything else, use the field below)",
                     )
                     other_emotion_tb = gr.Textbox(
-                        label="Other emotion word (not in AuK's list)",
-                        placeholder='e.g. "bemused" - automatically wrapped as a Timbre instruction instead',
+                        label="Other emotion word (not in AuK's list above)",
+                        placeholder='e.g. "bemused" - a single word for a momentary reaction, wrapped '
+                                     "as a Timbre instruction automatically (different from the "
+                                     "voice-description field above, which is for the character's "
+                                     "lasting accent/age/personality)",
                     )
                     pitch_dd = gr.Dropdown(["none"] + PITCH_OPTIONS, value="none", label="Pitch (semitones)")
                     speed_dd = gr.Dropdown(["none"] + SPEED_OPTIONS, value="none", label="Speed")
@@ -637,7 +659,7 @@ with gr.Blocks(title="Audio-demo pipeline studio") as demo:
             )
             apply_btn.click(
                 apply_edit,
-                inputs=[current_ref_audio, history_state, enhance_dd, emotion_dd, other_emotion_tb,
+                inputs=[current_ref_audio, history_state, enhance_dd, timbre_tb, emotion_dd, other_emotion_tb,
                         pitch_dd, speed_dd, custom_instruction],
                 outputs=[current_ref_audio, history_state, history_label, ref_transcript_box],
             )