Hugging Face
Search models, datasets, users...
Ctrl K
Models
Datasets
Spaces
Buckets
Docs
Enterprise
Pricing
Website
Tasks
HuggingChat
Collections
Languages
Organizations
Community
Blog
Posts
Daily Papers
Hardware
Learn
Discord
Forum
GitHub
Solutions
Team & Enterprise
Hugging Face PRO
Enterprise Support
Inference Providers
Inference Endpoints
Storage Buckets
Log In
Sign Up
Scicom-intl
/
Clearline-NC-CPU
Like
2
Follow
Scicom (MSC) Bhd
44
Audio-to-Audio
ONNX
speech-enhancement
noise-suppression
streaming
livekit
cpu
telephony
Model card
Files
Files and versions
xet
Community
Copy to bucket
main
Clearline-NC-CPU
27.6 MB
Ctrl+K
Ctrl+K
1 contributor
History:
13 commits
huseinzolkepliscicom
Card: shorter, in sections (models, use, results, CPU, limits); preprint pending. Weights and code unchanged
8d445c4
verified
about 16 hours ago
clearline
Card and a docstring: the fused graphs run two fifths of the plain graphs operators (78 of 189, 66 of 174), not a third; weights and graphs unchanged
about 17 hours ago
figures
2026-11-11: the plugins cost less (the VAD stream resamples in numpy and shares one inference between a session two streams; the noise filter ships fused graphs and keeps its buffers in place, which also fixes a slowdown at room rates other than 16 kHz); the weights are the same
about 17 hours ago
plain
2026-11-11: the plugins cost less (the VAD stream resamples in numpy and shares one inference between a session two streams; the noise filter ships fused graphs and keeps its buffers in place, which also fixes a slowdown at room rates other than 16 kHz); the weights are the same
about 17 hours ago
pytorch
Quality mode distilled from the GPU teacher (Clearline-NC-GPU): 256 units with per-bin levels and gains; ASR mode unchanged; renamed from Clearline-NC and opened
13 days ago
.gitattributes
Safe
1.52 kB
Clearline-NC v1 (2026-10-26): ONNX graphs, PyTorch weights, LiveKit plugin, model card
17 days ago
README.md
6.8 kB
Card: shorter, in sections (models, use, results, CPU, limits); preprint pending. Weights and code unchanged
about 16 hours ago
clearline_nc.onnx
5.27 MB
xet
2026-11-11: the plugins cost less (the VAD stream resamples in numpy and shares one inference between a session two streams; the noise filter ships fused graphs and keeps its buffers in place, which also fixes a slowdown at room rates other than 16 kHz); the weights are the same
about 17 hours ago
clearline_nc_asr.onnx
1.54 MB
xet
2026-11-11: the plugins cost less (the VAD stream resamples in numpy and shares one inference between a session two streams; the noise filter ships fused graphs and keeps its buffers in place, which also fixes a slowdown at room rates other than 16 kHz); the weights are the same
about 17 hours ago