Downloads · 30 days
0
dubbing-light-asd
blaze-voice-ai/dubbing-light-asd
dubbing-light-asd is a machine learning model from blaze-voice-ai. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.
Light-ASD finetuningTalkSet weights, from Junhua-Liao/Light-ASD (MIT). Mirrored because the upstream repo ships them in-tree and the service keeps that clone out of git, so this was the only copy on one machine.
Access
Public
Updated Sep 1, 2026
Repo size
4.2 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.model4.2 MB · 100%
At a glance
- License
- mit
- Access
- Public
- Created
- Sep 1, 2026
- Updated
- Sep 1, 2026
- SHA
- 94c290d0
Model card
From the Hugging Face model README
dubbing-light-asd
Light-ASD finetuning_TalkSet weights, from
Junhua-Liao/Light-ASD (MIT). Mirrored because the
upstream repo ships them in-tree and the service keeps that clone out of git, so this was the
only copy on one machine.
Measured on 100 labelled episodes of Chinese short drama, this model does not beat the
mouth-motion heuristic it was meant to replace: 54% coverage at 82.1% gender accuracy against
87% at 83.5%. Voice, heard across a whole speaker group, gets 91.4%. The face branch is
therefore off by default (FACE_DIAR=0); these weights exist so the measurement can be
reproduced.
Branches: main is what builds pin. dev for a retrained checkpoint.
- License
- mit
- Created
- Sep 1, 2026
- Updated
- Sep 1, 2026