Spaces:
Sleeping
Sleeping
File size: 4,585 Bytes
bc3f652 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 | """Generate speech fixtures with Qwen3-TTS and record their provenance."""
import argparse
import json
from pathlib import Path
import time
import numpy as np
import soundfile as sf
import torch
from huggingface_hub import snapshot_download
from qwen_tts import Qwen3TTSModel
def main():
parser = argparse.ArgumentParser()
parser.add_argument('--model-path')
parser.add_argument('--device', choices=['auto', 'cpu', 'cuda'], default='auto')
parser.add_argument('--model', help='Hugging Face model ID; inferred from local source.json when available')
parser.add_argument('--samples', type=Path, help='JSON list of name/text/language and speaker or instruct records')
parser.add_argument('--out', type=Path, default=Path('outputs/tts'))
parser.add_argument('--seed', type=int, default=42)
parser.add_argument('--only', help='Generate only the named sample')
parser.add_argument('--instruct', help='Override the voice description for every sample')
args = parser.parse_args()
args.out.mkdir(parents=True, exist_ok=True)
source_path = Path(args.model_path) / 'source.json' if args.model_path else None
source = json.loads(source_path.read_text()) if source_path and source_path.exists() else None
repo = args.model or (source or {}).get('repo') or 'Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice'
if source and args.model and source['repo'] != args.model:
raise ValueError('Requested model does not match local model provenance')
path = args.model_path or snapshot_download(repo, revision='5ecdb67327fd37bb2e042aab12ff7391903235d3' if repo == 'Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign' else None)
rows = json.loads(args.samples.read_text()) if args.samples else [
dict(name='zh', text='你好，今天的天气真好。', language='Chinese', speaker='Vivian'),
dict(name='en', text='Hello, can you hear the piano speaking?', language='English', speaker='Ryan'),
]
if args.only:
rows = [row for row in rows if row['name'] == args.only]
names = [row['name'] for row in rows]
if not names or len(set(names)) != len(names) or any(Path(n).name != n or n in ('', '.', '..') for n in names):
raise ValueError('Samples must have unique file stems, and the selection must not be empty')
torch.manual_seed(args.seed)
np.random.seed(args.seed)
device = ('cuda' if torch.cuda.is_available() else 'cpu') if args.device == 'auto' else args.device
dtype = torch.bfloat16 if device == 'cuda' else torch.float32
model = Qwen3TTSModel.from_pretrained(
path, device_map=device, dtype=dtype, attn_implementation='sdpa')
mode = model.model.tts_model_type
samples = []
for row in rows:
instruction = args.instruct if args.instruct is not None else row.get('instruct')
seed = row.get('seed', args.seed)
torch.manual_seed(seed)
np.random.seed(seed)
start = time.time()
if mode == 'voice_design':
if not instruction:
raise ValueError('VoiceDesign requires an instruct field or --instruct')
wavs, sr = model.generate_voice_design(
text=row['text'], language=row['language'], instruct=instruction, max_new_tokens=512)
elif mode == 'custom_voice':
if instruction and model.model.tts_model_size == '0b6':
raise ValueError('0.6B CustomVoice ignores instructions; use a supported 1.7B model')
wavs, sr = model.generate_custom_voice(
text=row['text'], language=row['language'], speaker=row['speaker'],
instruct=instruction, max_new_tokens=512)
else:
raise ValueError(f'Unsupported generation mode: {mode}')
sf.write(args.out / f"{row['name']}.wav", wavs[0], sr, subtype='FLOAT')
record = dict(row, seed=seed, sample_rate=sr, seconds=len(wavs[0]) / sr,
wall_seconds=time.time() - start)
if instruction:
record['instruct'] = instruction
if mode == 'voice_design':
record.pop('speaker', None)
samples.append(record)
print(record, flush=True)
manifest = dict(model=repo, snapshot=str(path), source=source, mode=mode,
seed=args.seed, seed_policy='Reset RNG before each sample',
device=device, dtype=str(dtype).removeprefix('torch.'), attention='sdpa', max_new_tokens=512, samples=samples)
(args.out / 'manifest.json').write_text(json.dumps(manifest, ensure_ascii=False, indent=2) + '\n')
if __name__ == '__main__':
main()
|