Configurable prompt seconds

app.py

@@ -8,13 +8,13 @@ import numpy as np
 
 base_url = "https://api.sandbox.deepgram.com/nlu"
 token_str = os.environ['DG_TOKEN']
-def tts_fn(text, prompt_audio, pitch_steps, inference_steps, inference_temperature):
+def tts_fn(text, prompt_audio, prompt_seconds, inference_steps, inference_temperature, pitch_steps):
     texts = [text]
     prompt_audio = np.reshape(prompt_audio[1], (1, 1, -1)).astype(np.float32, order='C') / 32768.0
     response = requests.post(
         f'{base_url}', 
         files=[('texts', ('texts', json.dumps(texts), 'application/json')), ('prompt_audio', ('prompt_audio', json.dumps(prompt_audio.tolist()), 'application/json'))], 
-        params={'synthesize': 'true', 'pitch_steps': int(pitch_steps), 'soundstorm_steps': inference_steps, 'temperature': inference_temperature},
+        params={'synthesize': 'true', 'pitch_steps': int(pitch_steps), 'soundstorm_steps': inference_steps, 'temperature': inference_temperature, 'prompt_seconds': prompt_seconds},
         headers={
             'Authorization': f'Token {token_str}'
         },
@@ -37,16 +37,17 @@ with app:
                 pangram = "The beige hue on the waters of the loch impressed all, including the French queen, before she heard that symphony again, just as young Arthur wanted."
                 cherry = "Your request has been processed and the audio is ready for playback."
                 textbox = gr.TextArea(label="Text", placeholder="Type a sentence here", value=cherry)
-                prompt_audio = gr.Audio(label="Prompt Audio (first 3 seconds of selection)", source='upload')
+                prompt_audio = gr.Audio(label="Prompt Audio", source='upload')
                 examples = gr.Examples(label='Sample Speakers', examples=demo_files, inputs=prompt_audio)
                 # speed = gr.Slider(minimum=0.0, maximum=2.0, value=1.1, step=0.1, label="Speed")
-                pitch_steps = gr.Slider(minimum=-24, maximum=24, value=0, step=1, label="Pitch Steps: 12 to an octave")
                 # variability = gr.Slider(minimum=0.0, maximum=1.0, value=0.7, step=0.1, label="Variability")
                 inference_steps = gr.Slider(minimum=1, maximum=32, value=1, step=1, label="Inference Steps: quality vs latency tradeoff. Results are sometimes unstable for values >1.")
                 inference_temperature = gr.Slider(minimum=0.0, maximum=1.0, value=0.9, step=0.05, label="Temperature: fidelity vs variability tradeoff")
+                prompt_seconds = gr.Slider(minimum=1.0, maximum=10.0, value=3.0, step=1.0, label="Use first N seconds of prompt audio")
+                pitch_steps = gr.Slider(minimum=-24, maximum=24, value=0, step=1, label="Pitch Steps: 12 to an octave")
 
             with gr.Column():
                 audio_output = gr.Audio(label="Output Audio", elem_id='tts-audio')
                 btn = gr.Button("Generate")
-                btn.click(tts_fn, inputs=[textbox, prompt_audio, pitch_steps, inference_steps, inference_temperature], outputs=[audio_output])
+                btn.click(tts_fn, inputs=[textbox, prompt_audio, prompt_seconds, inference_steps, inference_temperature, pitch_steps], outputs=[audio_output])
 app.launch()