Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,89 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
import outetts
|
| 3 |
+
from outetts.models.config import GenerationConfig
|
| 4 |
+
from huggingface_hub import hf_hub_download, login
|
| 5 |
+
import os
|
| 6 |
+
import numpy as np
|
| 7 |
+
|
| 8 |
+
login(token=os.environ.get("HF_TOKEN"))
|
| 9 |
+
|
| 10 |
+
model_path = hf_hub_download(
|
| 11 |
+
repo_id="Lyte/DarijaTTS",
|
| 12 |
+
filename="unsloth.Q8_0.gguf",
|
| 13 |
+
)
|
| 14 |
+
model_config = outetts.GGUFModelConfig_v2(
|
| 15 |
+
model_path=model_path,
|
| 16 |
+
tokenizer_path="Lyte/DarijaTTS",
|
| 17 |
+
#verbose=False
|
| 18 |
+
)
|
| 19 |
+
interface = outetts.InterfaceGGUF(model_version="0.3", cfg=model_config)
|
| 20 |
+
|
| 21 |
+
# TTS function
|
| 22 |
+
def tts(text, temperature, repetition_penalty):
|
| 23 |
+
try:
|
| 24 |
+
gen_cfg = GenerationConfig(
|
| 25 |
+
text=text,
|
| 26 |
+
temperature=temperature,
|
| 27 |
+
repetition_penalty=repetition_penalty,
|
| 28 |
+
max_length=4096,
|
| 29 |
+
speaker=None
|
| 30 |
+
)
|
| 31 |
+
|
| 32 |
+
output = interface.generate(config=gen_cfg)
|
| 33 |
+
|
| 34 |
+
output_path = "output.wav"
|
| 35 |
+
output.save(output_path)
|
| 36 |
+
|
| 37 |
+
return output_path
|
| 38 |
+
except Exception as e:
|
| 39 |
+
raise gr.Error(f"Error generating audio: {str(e)}")
|
| 40 |
+
|
| 41 |
+
with gr.Blocks() as demo:
|
| 42 |
+
gr.Markdown("# Darija TTS Demo (alpha)")
|
| 43 |
+
gr.Markdown("### Note: this is an early alpha which still needs training.")
|
| 44 |
+
|
| 45 |
+
with gr.Row():
|
| 46 |
+
text_input = gr.Textbox(
|
| 47 |
+
label="Enter text for TTS",
|
| 48 |
+
value="السلام كيداير لاباس عليك؟"
|
| 49 |
+
)
|
| 50 |
+
|
| 51 |
+
output_audio = gr.Audio(label="Generated Audio", type="filepath")
|
| 52 |
+
generate_btn = gr.Button("Generate Audio")
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
with gr.Row():
|
| 56 |
+
temperature = gr.Slider(
|
| 57 |
+
minimum=0.01,
|
| 58 |
+
maximum=1.0,
|
| 59 |
+
value=0.1,
|
| 60 |
+
step=0.01,
|
| 61 |
+
label="Temperature"
|
| 62 |
+
)
|
| 63 |
+
|
| 64 |
+
repetition_penalty = gr.Slider(
|
| 65 |
+
minimum=0.8,
|
| 66 |
+
maximum=1.5,
|
| 67 |
+
value=1.1,
|
| 68 |
+
step=0.05,
|
| 69 |
+
label="Repetition Penalty"
|
| 70 |
+
)
|
| 71 |
+
|
| 72 |
+
with gr.Row():
|
| 73 |
+
gr.Markdown("### Examples:")
|
| 74 |
+
gr.Examples(
|
| 75 |
+
[
|
| 76 |
+
["رمضان مبارك سعيد"],
|
| 77 |
+
["اشنو بان ليك فهادشي مزيان؟"],
|
| 78 |
+
["السلام كيداير لاباس عليك؟"],
|
| 79 |
+
["واش كاين شي جديد اليوم؟"]
|
| 80 |
+
],
|
| 81 |
+
inputs=[text_input])
|
| 82 |
+
|
| 83 |
+
generate_btn.click(
|
| 84 |
+
fn=tts,
|
| 85 |
+
inputs=[text_input, temperature, repetition_penalty],
|
| 86 |
+
outputs=output_audio
|
| 87 |
+
)
|
| 88 |
+
|
| 89 |
+
demo.launch(debug=True)
|