Update barks.py

barks.py

@@ -240,17 +240,53 @@ Oh-oh-oh, we're burnin' up the stage,
 Funk it up, we're livin' for the rage!"""
 
 def set_grunge_ballad_vocal_prompt():
-    return """[Verse 1, grunge ballad, male voice, soft]
+    return """[Verse 1 – Soft, Grunge Ballad, male voice]
 Shadows fall across my heart, I'm lost in the rain,
-Whispers of a broken dream, carry all my pain,
+Whispers of a broken dream, carry all my pain.
 Underneath the weight of time, I’m fading away,
 Searching for a spark to light another day.
 
-[Chorus, intense]
+[Chorus – Intense, male voice]
 Scream it out, let the silence break,
-Feel the fire, for my soul’s sake,
+Feel the fire, for my soul’s sake.
 Hold me now, through the endless night,
-In the dark, I’m reaching for the light!"""
+In the dark, I’m reaching for the light!
+
+[Verse 2 – Building Intensity, male voice]
+Cracks appear in my reflection, truth I can't deny,
+Memories like ghosts surround, no matter how I try.
+Each step forward feels like I'm walking through the past,
+Chasing echoes of a peace that never seems to last.
+
+[Chorus – Intensified, male voice]
+Scream it out, let the silence break,
+Feel the fire, for my soul’s sake.
+Hold me now, through the endless night,
+In the dark, I’m reaching for the light!
+
+[Bridge – Emotional Climax, male voice]
+I’ve been down this road before,
+Locked behind a closing door.
+But even in the blackest shade,
+A flicker of hope refuses to fade.
+
+[Verse 3 – Reflective, male voice]
+Rain-soaked streets and neon signs,
+Mark the path of these troubled times.
+Yet amidst the storm and strife,
+I find fragments of a former life.
+
+[Chorus – Final, Powerful, male voice]
+Scream it out, let the silence break,
+Feel the fire, for my soul’s sake.
+Hold me now, through the endless night,
+In the dark, I’m reaching for the light!
+
+[Outro – Soft, Resolute, male voice]
+Though shadows linger and nights are long,
+Within my soul, I find a song.
+A melody of hope, burning bright,
+Guiding me onward, into the light."""
 
 def set_indie_pop_vocal_prompt():
     return """[Verse 1, indie pop, female voice]
@@ -276,13 +312,24 @@ def apply_fade(segment, fade_in_duration=1000, fade_out_duration=1000):
     segment = segment.fade_out(fade_out_duration)
     return segment
 
-def generate_vocals(vocal_prompt: str, total_duration: int):
+def generate_vocals(vocal_prompt: str, total_duration: int, speaker_preset: str):
     global bark_model, bark_processor
     if not vocal_prompt.strip():
         return None, "⚠️ Please enter a valid vocal prompt!"
     
     try:
         print("Generating vocals with Bark...")
+        # Apply speaker preset if specified
+        if speaker_preset != "default":
+            vocal_prompt = f"[{speaker_preset}] {vocal_prompt}"
+        
+        # Check token length
+        tokens = bark_processor.tokenize(vocal_prompt)
+        token_ids = bark_processor.convert_tokens_to_ids(tokens)
+        if len(token_ids) > 512:
+            print("WARNING: Vocal prompt exceeds 512 tokens; truncating to avoid errors.")
+            vocal_prompt = bark_processor.decode(token_ids[:512])
+        
         # Process vocal prompt with attention mask, relying on default padding
         inputs = bark_processor(
             vocal_prompt,
@@ -323,7 +370,7 @@ def generate_vocals(vocal_prompt: str, total_duration: int):
         return None, f"❌ Vocal generation failed: {e}"
 
 # 7) GENERATION & I/O FUNCTIONS
-def generate_music(instrumental_prompt: str, vocal_prompt: str, cfg_scale: float, top_k: int, top_p: float, temperature: float, total_duration: int, chunk_duration: int, crossfade_duration: int, bpm: int, drum_beat: str, synthesizer: str, rhythmic_steps: str, bass_style: str, guitar_style: str):
+def generate_music(instrumental_prompt: str, vocal_prompt: str, cfg_scale: float, top_k: int, top_p: float, temperature: float, total_duration: int, chunk_duration: int, crossfade_duration: int, bpm: int, drum_beat: str, synthesizer: str, rhythmic_steps: str, bass_style: str, guitar_style: str, speaker_preset: str):
     global musicgen_model
     if not instrumental_prompt.strip():
         return None, "⚠️ Please enter a valid instrumental prompt!"
@@ -399,7 +446,7 @@ def generate_music(instrumental_prompt: str, vocal_prompt: str, cfg_scale: float
 
         # Generate vocals if provided
         if vocal_prompt.strip():
-            vocal_segment, vocal_status = generate_vocals(vocal_prompt, total_duration)
+            vocal_segment, vocal_status = generate_vocals(vocal_prompt, total_duration, speaker_preset)
             if vocal_segment is None:
                 return None, vocal_status
             print("Mixing vocals with instrumental...")
@@ -430,7 +477,7 @@ def generate_music(instrumental_prompt: str, vocal_prompt: str, cfg_scale: float
 
 # Function to clear inputs
 def clear_inputs():
-    return "", "", 3.0, 250, 0.9, 1.0, 30, 5, 1000, 120, "none", "none", "none", "none", "none"
+    return "", "", 3.0, 250, 0.9, 1.0, 30, 5, 1000, 120, "none", "none", "none", "none", "none", "default"
 
 # 8) CUSTOM CSS
 css = """
@@ -552,6 +599,13 @@ with gr.Blocks(css=css) as demo:
             funk_vocal_btn = gr.Button("Upbeat Funk Rock 🎵", elem_classes="vocal-btn")
             grunge_vocal_btn = gr.Button("Grunge Ballad 🎸", elem_classes="vocal-btn")
             indie_vocal_btn = gr.Button("Indie Pop 🎤", elem_classes="vocal-btn")
+        
+        speaker_preset = gr.Dropdown(
+            label="Bark Speaker Preset 🎙️",
+            choices=["default", "v2/en_speaker_6", "v2/en_speaker_9"],
+            value="default",
+            info="Select a speaker preset to influence vocal style (e.g., male or female voice). Note: Bark is primarily a speech model; singing styles are limited and best controlled via prompt cues (e.g., [soft], [intense])."
+        )
     
     with gr.Column(elem_classes="settings-container"):
         gr.Markdown("### ⚙️ API Settings")
@@ -680,13 +734,13 @@ with gr.Blocks(css=css) as demo:
     indie_vocal_btn.click(set_indie_pop_vocal_prompt, inputs=[], outputs=vocal_prompt)
     gen_btn.click(
         generate_music,
-        inputs=[instrumental_prompt, vocal_prompt, cfg_scale, top_k, top_p, temperature, total_duration, chunk_duration, crossfade_duration, bpm, drum_beat, synthesizer, rhythmic_steps, bass_style, guitar_style],
+        inputs=[instrumental_prompt, vocal_prompt, cfg_scale, top_k, top_p, temperature, total_duration, chunk_duration, crossfade_duration, bpm, drum_beat, synthesizer, rhythmic_steps, bass_style, guitar_style, speaker_preset],
         outputs=[out_audio, status]
     )
     clr_btn.click(
         clear_inputs,
         inputs=None,
-        outputs=[instrumental_prompt, vocal_prompt, cfg_scale, top_k, top_p, temperature, total_duration, chunk_duration, crossfade_duration, bpm, drum_beat, synthesizer, rhythmic_steps, bass_style, guitar_style]
+        outputs=[instrumental_prompt, vocal_prompt, cfg_scale, top_k, top_p, temperature, total_duration, chunk_duration, crossfade_duration, bpm, drum_beat, synthesizer, rhythmic_steps, bass_style, guitar_style, speaker_preset]
     )
 
 # 10) TURN OFF OPENAPI/DOCS