Spaces:
Sleeping
Sleeping
| import gradio as gr | |
| import torch | |
| from parler_tts import ParlerTTSForConditionalGeneration | |
| from transformers import AutoTokenizer, set_seed | |
| import numpy as np | |
| # Set device | |
| device = "cuda:0" if torch.cuda.is_available() else "cpu" | |
| # Load model and tokenizer | |
| model = ParlerTTSForConditionalGeneration.from_pretrained("TArtx/parler-tts-mini-v1-finetuned-12").to(device) | |
| tokenizer = AutoTokenizer.from_pretrained("parler-tts/parler-tts-mini-v1") | |
| # Constants | |
| SAMPLE_RATE = model.config.sampling_rate | |
| SEED = 42 | |
| # Default inputs | |
| default_text = "This is a demonstration of my ability to convert written words into spoken language, seamlessly and naturally. As a text-to-speech model, my goal is to sound as clear and engaging as a human, making sure every word I say leaves an impression." | |
| default_description = "moderate speed, very clear, monotone, wonderful speech quality" | |
| # TTS generation function | |
| def gen_tts(text, description): | |
| try: | |
| # Set seed for reproducibility | |
| set_seed(SEED) | |
| # Prepare inputs | |
| input_ids = tokenizer(description.strip(), return_tensors="pt").input_ids.to(device) | |
| prompt_input_ids = tokenizer(text.strip(), return_tensors="pt").input_ids.to(device) | |
| # Generate audio | |
| generation = model.generate( | |
| input_ids=input_ids, | |
| prompt_input_ids=prompt_input_ids, | |
| do_sample=True, | |
| temperature=0.7 | |
| ) | |
| # Convert to numpy array | |
| audio_arr = generation.cpu().numpy().squeeze() | |
| # Normalize audio | |
| if np.max(np.abs(audio_arr)) > 0: | |
| audio_arr = audio_arr / np.max(np.abs(audio_arr)) | |
| audio_arr = (audio_arr * np.iinfo(np.int16).max).astype(np.int16) | |
| else: | |
| # Fallback to white noise if generation fails | |
| audio_arr = np.random.randint(-32768, 32767, SAMPLE_RATE * 10, dtype=np.int16) | |
| return SAMPLE_RATE, audio_arr | |
| except Exception as e: | |
| print(f"Error in TTS generation: {str(e)}") | |
| # Return white noise as fallback | |
| return SAMPLE_RATE, np.random.randint(-32768, 32767, SAMPLE_RATE * 10, dtype=np.int16) | |
| # Gradio interface | |
| with gr.Blocks() as block: | |
| gr.Markdown( | |
| """ | |
| ## Parler-TTS 🗣️ | |
| Parler-TTS is a training and inference library for high-fidelity text-to-speech (TTS) models. This demo uses the Mini v1 model. | |
| """ | |
| ) | |
| with gr.Row(): | |
| with gr.Column(): | |
| input_text = gr.Textbox(label="Input Text", lines=2, value=default_text, elem_id="input_text") | |
| description = gr.Textbox(label="Description", lines=2, value=default_description, elem_id="input_description") | |
| run_button = gr.Button("Generate Audio", variant="primary") | |
| with gr.Column(): | |
| audio_out = gr.Audio(label="Parler-TTS generation", type="numpy", elem_id="audio_out") | |
| inputs = [input_text, description] | |
| outputs = audio_out | |
| run_button.click(fn=gen_tts, inputs=inputs, outputs=outputs) | |
| # Launch the interface | |
| block.launch(debug=True) |