Improvements to synthesis via filelist and simple filenames - #853
Conversation
fix filenames for vocoder matching
add simple filename synthesis fixes #716
Changed Files
|
update to pytest
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #853 +/- ##
==========================================
+ Coverage 85.21% 85.22% +0.01%
==========================================
Files 51 51
Lines 4706 4710 +4
Branches 699 700 +1
==========================================
+ Hits 4010 4014 +4
Misses 535 535
Partials 161 161 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
|
Generally looks good. Effective solution. There's a test that fails in CI in fs2, and comments in both sub PRs. |
minor fixes suggested by @joanise for synthesis improvements
joanise
left a comment
There was a problem hiding this comment.
Looks good, thanks!
Conflicting submodule changes pushed, now you'll have to merge and/or rebase the submodules and redo a clean submodule update in this top-level PR.
PR Goal?
This PR does a couple small improvements to the synthesis operation in FS2 and STTS2. First, it implements the suggestion in #716 to allow simple basename synthesis when given a filelist. If you pass --simple-filenames then you just get the basename.wav
it also implements synthesizing with a filelist for STTS2 and fixes a small bug in the demo. All tested on trained models.
Assisted by Claude Sonnet 5
Fixes?
#716
Feedback sought?
rubber stamping, confirmation that the simple-filenames feature works as intended.
Priority?
medium
Tests added?
Some basic tests added
How to test?
Should work with existing models
Confidence?
pretty high
Version change?
Related PRs?
EveryVoiceTTS/FastSpeech2_lightning#154
EveryVoiceTTS/StyleTTS2#24