Implement Prosody metrics for TTS evaluation - #16031

Open

Edresson wants to merge 10 commits into

NVIDIA-NeMo:mainfrom

Edresson:NeMo_main_08_02_new_metrics

Commits on Aug 7, 2026

Support TTS speaker id format on SpeakerFilter class
Edresson
committed
4cb90ef
Add prosody evaluation metrics
Edresson
committed
1b39526
Clean up emotion encoder CLI
Edresson
committed
40bef8e
Simplify prosody metric defaults
Edresson
committed
2a8b8f1
Use default prosody emotion embedding
Edresson
committed
07c8740
Remove emotion encoder CLI
Edresson
committed
ad26f40
Fix black formatting
Edresson
committed
c8b9edc
Update speaker filter test expectation
Edresson
committed
c912db2

Commits on Aug 10, 2026

Add generated-to-ground-truth audio CER metric
Edresson
committed
f7db3e5
Add generated-to-ground-truth audio WER metric
Edresson
committed
75c5827