SOTA Open Source TTS
from typing import Callable
import torch
from loguru import logger
from fish_speech.models.dac.modded_dac import DAC
class VQManager:
def __init__(self):
# Make Pylance happy (attribut/method not defined...)
self.decoder_model: DAC
self.load_audio: Callable
def decode_vq_tokens(self, codes):
logger.info(f"VQ features: {codes.shape}")
if isinstance(self.decoder_model, DAC):
# Drama 3 hands over its codes on the CPU.
codes = codes.to(self.decoder_model.device)
return self.decoder_model.from_indices(codes[None])[0].squeeze()
raise ValueError(f"Unknown model type: {type(self.decoder_model)}")
def encode_reference(self, reference_audio, enable_reference_audio):
if enable_reference_audio and reference_audio is not None:
# Load audios, and prepare basic info here
if hasattr(self.decoder_model, "spec_transform"):
sample_rate = self.decoder_model.spec_transform.sample_rate
else:
sample_rate = self.decoder_model.sample_rate
reference_audio_content = self.load_audio(reference_audio, sample_rate)
audios = torch.from_numpy(reference_audio_content).to(
device=self.decoder_model.device,
dtype=next(self.decoder_model.parameters()).dtype,
)[None, None, :]
audio_lengths = torch.tensor(
[audios.shape[2]], device=self.decoder_model.device, dtype=torch.long
)
logger.info(
f"Loaded audio with {audios.shape[2] / sample_rate:.2f} seconds"
)
# VQ Encoder
if isinstance(self.decoder_model, DAC):
prompt_tokens = self.decoder_model.encode(audios, audio_lengths)[0][0]
logger.info(f"Encoded prompt: {prompt_tokens.shape}")
else:
raise ValueError(f"Unknown model type: {type(self.decoder_model)}")
else:
prompt_tokens = None
logger.info("No reference audio provided")
return prompt_tokens