tortoise-tts/tortoise/read.py

94 lines
4.9 KiB
Python
Raw Permalink Normal View History

2022-04-11 01:29:42 +00:00
import argparse
import os
from time import time
2022-04-11 01:29:42 +00:00
import torch
import torchaudio
from api import TextToSpeech, MODELS_DIR
from utils.audio import load_audio, load_voices
2022-05-13 11:02:17 +00:00
from utils.text import split_and_recombine_text
2022-04-11 01:29:42 +00:00
2022-04-21 22:06:43 +00:00
2022-04-11 01:29:42 +00:00
if __name__ == '__main__':
parser = argparse.ArgumentParser()
2022-05-03 02:56:28 +00:00
parser.add_argument('--textfile', type=str, help='A file containing the text to read.', default="tortoise/data/riding_hood.txt")
2022-04-15 14:26:11 +00:00
parser.add_argument('--voice', type=str, help='Selects the voice to use for generation. See options in voices/ directory (and add your own!) '
2022-04-26 02:07:07 +00:00
'Use the & character to join two voices together. Use a comma to perform inference on multiple voices.', default='pat')
2022-05-03 03:37:39 +00:00
parser.add_argument('--output_path', type=str, help='Where to store outputs.', default='results/longform/')
2022-04-20 23:24:09 +00:00
parser.add_argument('--preset', type=str, help='Which voice preset to use.', default='standard')
2022-04-26 02:05:21 +00:00
parser.add_argument('--regenerate', type=str, help='Comma-separated list of clip numbers to re-generate, or nothing.', default=None)
parser.add_argument('--candidates', type=int, help='How many output candidates to produce per-voice. Only the first candidate is actually used in the final product, the others can be used manually.', default=1)
parser.add_argument('--model_dir', type=str, help='Where to find pretrained model checkpoints. Tortoise automatically downloads these to .models, so this'
2022-05-19 16:37:57 +00:00
'should only be specified if you have custom checkpoints.', default=MODELS_DIR)
parser.add_argument('--seed', type=int, help='Random seed which can be used to reproduce results.', default=None)
parser.add_argument('--produce_debug_state', type=bool, help='Whether or not to produce debug_state.pth, which can aid in reproducing problems. Defaults to true.', default=True)
2022-04-11 01:29:42 +00:00
args = parser.parse_args()
2022-05-03 00:00:57 +00:00
tts = TextToSpeech(models_dir=args.model_dir)
2022-04-11 01:29:42 +00:00
2022-04-15 14:26:11 +00:00
outpath = args.output_path
selected_voices = args.voice.split(',')
2022-04-26 02:05:21 +00:00
regenerate = args.regenerate
if regenerate is not None:
regenerate = [int(e) for e in regenerate.split(',')]
2022-05-12 17:24:55 +00:00
# Process text
with open(args.textfile, 'r', encoding='utf-8') as f:
text = ' '.join([l for l in f.readlines()])
if '|' in text:
print("Found the '|' character in your text, which I will use as a cue for where to split it up. If this was not"
"your intent, please remove all '|' characters from the input.")
texts = text.split('|')
else:
texts = split_and_recombine_text(text)
seed = int(time()) if args.seed is None else args.seed
2022-04-15 14:26:11 +00:00
for selected_voice in selected_voices:
voice_outpath = os.path.join(outpath, selected_voice)
os.makedirs(voice_outpath, exist_ok=True)
if '&' in selected_voice:
voice_sel = selected_voice.split('&')
else:
voice_sel = [selected_voice]
2022-04-11 01:29:42 +00:00
voice_samples, conditioning_latents = load_voices(voice_sel)
all_parts = []
2022-04-15 14:26:11 +00:00
for j, text in enumerate(texts):
2022-04-26 02:05:21 +00:00
if regenerate is not None and j not in regenerate:
all_parts.append(load_audio(os.path.join(voice_outpath, f'{j}.wav'), 24000))
continue
gen = tts.tts_with_preset(text, voice_samples=voice_samples, conditioning_latents=conditioning_latents,
preset=args.preset, k=args.candidates, use_deterministic_seed=seed)
if args.candidates == 1:
gen = gen.squeeze(0).cpu()
torchaudio.save(os.path.join(voice_outpath, f'{j}.wav'), gen, 24000)
else:
candidate_dir = os.path.join(voice_outpath, str(j))
os.makedirs(candidate_dir, exist_ok=True)
for k, g in enumerate(gen):
torchaudio.save(os.path.join(candidate_dir, f'{k}.wav'), g.squeeze(0).cpu(), 24000)
gen = gen[0].squeeze(0).cpu()
all_parts.append(gen)
if args.candidates == 1:
full_audio = torch.cat(all_parts, dim=-1)
torchaudio.save(os.path.join(voice_outpath, 'combined.wav'), full_audio, 24000)
2022-04-11 01:29:42 +00:00
if args.produce_debug_state:
os.makedirs('debug_states', exist_ok=True)
dbg_state = (seed, texts, voice_samples, conditioning_latents)
torch.save(dbg_state, f'debug_states/read_debug_{selected_voice}.pth')
# Combine each candidate's audio clips.
if args.candidates > 1:
audio_clips = []
for candidate in range(args.candidates):
for line in range(len(texts)):
wav_file = os.path.join(voice_outpath, str(line), f"{candidate}.wav")
audio_clips.append(load_audio(wav_file, 24000))
audio_clips = torch.cat(audio_clips, dim=-1)
torchaudio.save(os.path.join(voice_outpath, f"combined_{candidate:02d}.wav"), audio_clips, 24000)
audio_clips = []