Medix-AI/coqui-tts
Medix-AI
/
coqui-tts
mirror of https://github.com/coqui-ai/TTS.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #3058 from coqui-ai/spkr_enc_3020 

fixed bugs in fastpitch tts synthesis
This commit is contained in:
Eren Gölge 2023-10-14 11:40:31 +02:00 committed by GitHub
parent 99635193f5 ffddf10458
commit b25d96ecee
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 3 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
TTS/tts/models/forward_tts.py
View File

@ -396,6 +396,7 @@ class ForwardTTS(BaseTTS):
- g: :math:`(B, C)` - g: :math:`(B, C)`
""" """
if hasattr(self, "emb_g"): if hasattr(self, "emb_g"):
g = g.type(torch.LongTensor)
g = self.emb_g(g) # [B, C, 1] g = self.emb_g(g) # [B, C, 1]
if g is not None: if g is not None:
g = g.unsqueeze(-1) g = g.unsqueeze(-1)
@ -683,9 +684,10 @@ class ForwardTTS(BaseTTS):
# encoder pass # encoder pass
o_en, x_mask, g, _ = self._forward_encoder(x, x_mask, g) o_en, x_mask, g, _ = self._forward_encoder(x, x_mask, g)
# duration predictor pass # duration predictor pass
o_dr_log = self.duration_predictor(o_en, x_mask) o_dr_log = self.duration_predictor(o_en.squeeze(), x_mask)
o_dr = self.format_durations(o_dr_log, x_mask).squeeze(1) o_dr = self.format_durations(o_dr_log, x_mask).squeeze(1)
y_lengths = o_dr.sum(1) y_lengths = o_dr.sum(1)
# pitch predictor pass # pitch predictor pass
o_pitch = None o_pitch = None
if self.args.use_pitch: if self.args.use_pitch: