Spaces:
Paused
Paused
Delete app-backup.py
Browse files- app-backup.py +0 -117
app-backup.py
DELETED
|
@@ -1,117 +0,0 @@
|
|
| 1 |
-
import os
|
| 2 |
-
import gradio as gr
|
| 3 |
-
import outetts
|
| 4 |
-
from outetts.version.v2.interface import _DEFAULT_SPEAKERS
|
| 5 |
-
import torch
|
| 6 |
-
import spaces
|
| 7 |
-
|
| 8 |
-
def get_available_speakers():
|
| 9 |
-
speakers = list(_DEFAULT_SPEAKERS.keys())
|
| 10 |
-
return speakers
|
| 11 |
-
|
| 12 |
-
@spaces.GPU
|
| 13 |
-
def generate_tts(text, temperature, repetition_penalty, speaker_selection, reference_audio):
|
| 14 |
-
model_config = outetts.HFModelConfig_v2(
|
| 15 |
-
model_path="OuteAI/OuteTTS-0.3-1B",
|
| 16 |
-
tokenizer_path="OuteAI/OuteTTS-0.3-1B",
|
| 17 |
-
dtype=torch.bfloat16,
|
| 18 |
-
device="cuda"
|
| 19 |
-
)
|
| 20 |
-
interface = outetts.InterfaceHF(model_version="0.3", cfg=model_config)
|
| 21 |
-
|
| 22 |
-
try:
|
| 23 |
-
if reference_audio:
|
| 24 |
-
speaker = interface.create_speaker(reference_audio)
|
| 25 |
-
elif speaker_selection and speaker_selection != "None":
|
| 26 |
-
speaker = interface.load_default_speaker(speaker_selection)
|
| 27 |
-
else:
|
| 28 |
-
speaker = None
|
| 29 |
-
|
| 30 |
-
gen_cfg = outetts.GenerationConfig(
|
| 31 |
-
text=text,
|
| 32 |
-
temperature=temperature,
|
| 33 |
-
repetition_penalty=repetition_penalty,
|
| 34 |
-
max_length=4096,
|
| 35 |
-
speaker=speaker,
|
| 36 |
-
)
|
| 37 |
-
output = interface.generate(config=gen_cfg)
|
| 38 |
-
|
| 39 |
-
if output.audio is None:
|
| 40 |
-
raise ValueError("Model failed to generate audio. This may be due to input length constraints or early EOS token.")
|
| 41 |
-
|
| 42 |
-
output_path = "output.wav"
|
| 43 |
-
output.save(output_path)
|
| 44 |
-
return output_path, None
|
| 45 |
-
except Exception as e:
|
| 46 |
-
return None, str(e)
|
| 47 |
-
|
| 48 |
-
with gr.Blocks(theme="Yntec/HaleyCH_Theme_Orange") as demo:
|
| 49 |
-
gr.Markdown("# Voice Clone Multilingual TTS")
|
| 50 |
-
|
| 51 |
-
error_box = gr.Textbox(label="Error Messages", visible=False)
|
| 52 |
-
|
| 53 |
-
with gr.Row():
|
| 54 |
-
with gr.Column(scale=1):
|
| 55 |
-
text_input = gr.Textbox(
|
| 56 |
-
label="Text to Synthesize",
|
| 57 |
-
placeholder="Enter text here...",
|
| 58 |
-
lines=8
|
| 59 |
-
)
|
| 60 |
-
|
| 61 |
-
submit_button = gr.Button("Generate Speech")
|
| 62 |
-
|
| 63 |
-
with gr.Column(scale=1):
|
| 64 |
-
audio_output = gr.Audio(
|
| 65 |
-
label="Generated Audio",
|
| 66 |
-
type="filepath"
|
| 67 |
-
)
|
| 68 |
-
|
| 69 |
-
with gr.Group():
|
| 70 |
-
speaker_dropdown = gr.Dropdown(
|
| 71 |
-
choices=get_available_speakers(),
|
| 72 |
-
value="en_male_1",
|
| 73 |
-
label="Speaker Selection"
|
| 74 |
-
)
|
| 75 |
-
|
| 76 |
-
temperature = gr.Slider(
|
| 77 |
-
0.1, 1.0,
|
| 78 |
-
value=0.1,
|
| 79 |
-
label="Temperature (lower = more stable tone, higher = more expressive)"
|
| 80 |
-
)
|
| 81 |
-
|
| 82 |
-
repetition_penalty = gr.Slider(
|
| 83 |
-
0.5, 2.0,
|
| 84 |
-
value=1.1,
|
| 85 |
-
label="Repetition Penalty"
|
| 86 |
-
)
|
| 87 |
-
|
| 88 |
-
reference_audio = gr.Audio(
|
| 89 |
-
label="Reference Audio (for voice cloning)",
|
| 90 |
-
type="filepath"
|
| 91 |
-
)
|
| 92 |
-
|
| 93 |
-
gr.Markdown("""
|
| 94 |
-
### Voice Cloning Guidelines:
|
| 95 |
-
- Use around 7-10 seconds of clear, noise-free audio
|
| 96 |
-
- For transcription interface will use Whisper turbo to transcribe the audio file
|
| 97 |
-
- Longer audio clips will reduce maximum output length
|
| 98 |
-
- Custom speaker overrides speaker selection
|
| 99 |
-
""")
|
| 100 |
-
|
| 101 |
-
submit_button.click(
|
| 102 |
-
fn=generate_tts,
|
| 103 |
-
inputs=[
|
| 104 |
-
text_input,
|
| 105 |
-
temperature,
|
| 106 |
-
repetition_penalty,
|
| 107 |
-
speaker_dropdown,
|
| 108 |
-
reference_audio,
|
| 109 |
-
],
|
| 110 |
-
outputs=[audio_output, error_box]
|
| 111 |
-
).then(
|
| 112 |
-
fn=lambda x: gr.update(visible=bool(x)),
|
| 113 |
-
inputs=[error_box],
|
| 114 |
-
outputs=[error_box]
|
| 115 |
-
)
|
| 116 |
-
|
| 117 |
-
demo.launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|