deepgeek/CosyVoice
2
0
Fork 0
mirror of https://github.com/FunAudioLLM/CosyVoice.git synced 2026-02-04 17:39:25 +08:00
Code Issues Packages Projects Releases Wiki Activity

add vc code

Browse Source
This commit is contained in:
lyuxiang.lx
2024-09-26 10:49:22 +08:00
parent ed87445540
commit 49015f63e6
7 changed files with 43 additions and 216 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
cosyvoice/flow/flow.py
View File

@@ -124,7 +124,7 @@ class MaskedDiffWithXvec(torch.nn.Module):
# text encode
h, h_lengths = self.encoder(token, token_len)
h = self.encoder_proj(h)
mel_len1, mel_len2 = prompt_feat.shape[1], int(token_len2 / 50 * 22050 / 256)
mel_len1, mel_len2 = prompt_feat.shape[1], int(token_len2 / self.input_frame_rate * 22050 / 256)
h, h_lengths = self.length_regulator.inference(h[:, :token_len1], h[:, token_len1:], mel_len1, mel_len2)
# get conditions
@@ -132,7 +132,6 @@ class MaskedDiffWithXvec(torch.nn.Module):
conds[:, :mel_len1] = prompt_feat
conds = conds.transpose(1, 2)
# mask = (~make_pad_mask(feat_len)).to(h)
mask = (~make_pad_mask(torch.tensor([mel_len1 + mel_len2]))).to(h)
feat = self.decoder(
mu=h.transpose(1, 2).contiguous(),