| """Run the standalone public DuDE release.""" |
| import argparse |
| import json |
| from pathlib import Path |
| from dude_tts import DuDE |
|
|
| if __name__ == '__main__': |
| p = argparse.ArgumentParser(description=__doc__) |
| p.add_argument('--model', default='penguinfish1688/duplexdataengine') |
| p.add_argument('--xml', required=True, help='Path to a dialogue XML file') |
| p.add_argument('--voice-a', default='voice_1', help='Speaker A reference audio path or included voice name') |
| p.add_argument('--voice-b', default='voice_2', help='Speaker B reference audio path or included voice name') |
| p.add_argument('--output', default='duplex.wav') |
| p.add_argument('--seed', type=int, default=20260922) |
| p.add_argument('--max-seconds', type=float, default=150.) |
| p.add_argument('--device', default='cuda') |
| args = p.parse_args() |
| model = DuDE.from_pretrained(args.model, device=args.device) |
| result = model.generate(Path(args.xml).read_text(), args.voice_a, args.voice_b, |
| seed=args.seed, max_seconds=args.max_seconds) |
| result.save(args.output) |
| print(json.dumps(dict(output=str(Path(args.output).resolve()), sample_rate=result.sample_rate, |
| eos=result.eos, frames=result.frames))) |
|
|
