QuickTalk weights for OpenTalking
This repository contains the QuickTalk runtime files used by OpenTalking.
Files
| Path | SHA256 | Notes |
|---|---|---|
quicktalk.pth |
fc8a7ea025c99a471ef00738874be5ecb6b5dfaf88ff6a1255a5d45a05d73001 |
TorchScript QuickTalk model; verified with torch.jit.load. |
repair.npy |
9ea50edde851bf3b12aa22d67b6f0db4f2930f3d9b7b3febcbd383e14117bfca |
QuickTalk repair asset. |
chinese-hubert-large/config.json |
8511d73054ac289ef47a527efdfd6738d2cb60f69f2973fdc9277492d9ff854b |
HuBERT config matching the validated OpenTalking QuickTalk runtime. |
chinese-hubert-large/preprocessor_config.json |
6334d6e0c5f2084c9a99b85ddff243cbc79dbaa4aa790bcddf8c41c496fab6fb |
HuBERT feature extractor config matching the validated OpenTalking QuickTalk runtime. |
chinese-hubert-large/pytorch_model.bin |
9cf43abec3f0410ad6854afa4d376c69ccb364b48ddddfd25c4c5aa16398eab0 |
HuBERT checkpoint matching facebook/hubert-large-ls960-ft, Apache-2.0 License. |
The HuBERT directory mirrors the file combination validated with OpenTalking QuickTalk. Its pytorch_model.bin matches the public facebook/hubert-large-ls960-ft checkpoint; the config files are kept as the validated runtime configs. Please review the upstream model cards and licenses before redistribution.
Download
hf download datascale-ai/quicktalk \
quicktalk.pth \
repair.npy \
chinese-hubert-large/config.json \
chinese-hubert-large/preprocessor_config.json \
chinese-hubert-large/pytorch_model.bin \
--local-dir models/quicktalk/checkpoints
OpenTalking still requires InsightFace buffalo_l files under models/quicktalk/checkpoints/auxiliary/models/buffalo_l/.