VibeVoice
Collection
Frontier Text-to-Speech Models https://microsoft.github.io/VibeVoice/ ā¢ 8 items ā¢ Updated
ā¢ 211
Joshua Chris
KrisKale45
AI & ML interests
None yet
Recent Activity
upvoted a collection 4 days ago
VibeVoice liked
a model 5 days ago
janhq/Jan-code-4b-gguf liked
a model 11 days ago
Jackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled Organizations
None yet
upvoted a collection 4 days ago
liked a model 5 days ago
liked a model 11 days ago
liked a model 15 days ago
reacted to robtacconelli's post with šš 15 days ago
Post
3632
š Nacrith: a 135M model that out-compresses everything on natural language
What if a tiny LM could compress english text better than _every_ compressor out there ā classical or neural, small or large?
Nacrith pairs SmolLM2-135M with an ensemble of online predictors and high-precision arithmetic coding.
What's inside
The standard LLM+arithmetic coding approach wastes ~75% of CDF precision on large vocabularies. Our CDF-24 fix alone recovers 0.5 bpb. On top: a token N-gram that skips the GPU on predictable tokens, an adaptive bias head, llama.cpp backend (7Ć faster than PyTorch), multi-GPU parallel compression, and a binary file format (NC06) ā the first LLM-based binary compressor we know of.
Runs on a GTX 1050 Ti. ~500 MB weights, ~1.2 GB VRAM per worker.
š» Code: https://github.com/robtacconelli/Nacrith-GPU
ā Space: robtacconelli/Nacrith-GPU
š Paper: Nacrith: Neural Lossless Compression via Ensemble Context Modeling and High-Precision CDF Coding (2602.19626)
Try it, break it, share your results ā all feedback welcome. ā on the repo appreciated!
Results across all systems we tested:
- alice29.txt ā 0.918 bpb (ā44% vs CMIX, ā20% vs ts_zip) ā below the 2nd-order Shannon entropy bound
- enwik8 (100 MB) ā 0.9389 bpb (ā8% vs FineZip/LLMZip's 8B model, ā15% vs ts_zip)
- Unseen text ā 0.723 bpb on a doc published after training cutoff ā no memorization, 26% better than FineZip/LLMZip on the same model
SmolLM2-135M by
HuggingFaceTB
What if a tiny LM could compress english text better than _every_ compressor out there ā classical or neural, small or large?
Nacrith pairs SmolLM2-135M with an ensemble of online predictors and high-precision arithmetic coding.
What's inside
The standard LLM+arithmetic coding approach wastes ~75% of CDF precision on large vocabularies. Our CDF-24 fix alone recovers 0.5 bpb. On top: a token N-gram that skips the GPU on predictable tokens, an adaptive bias head, llama.cpp backend (7Ć faster than PyTorch), multi-GPU parallel compression, and a binary file format (NC06) ā the first LLM-based binary compressor we know of.
Runs on a GTX 1050 Ti. ~500 MB weights, ~1.2 GB VRAM per worker.
š» Code: https://github.com/robtacconelli/Nacrith-GPU
ā Space: robtacconelli/Nacrith-GPU
š Paper: Nacrith: Neural Lossless Compression via Ensemble Context Modeling and High-Precision CDF Coding (2602.19626)
Try it, break it, share your results ā all feedback welcome. ā on the repo appreciated!
Results across all systems we tested:
- alice29.txt ā 0.918 bpb (ā44% vs CMIX, ā20% vs ts_zip) ā below the 2nd-order Shannon entropy bound
- enwik8 (100 MB) ā 0.9389 bpb (ā8% vs FineZip/LLMZip's 8B model, ā15% vs ts_zip)
- Unseen text ā 0.723 bpb on a doc published after training cutoff ā no memorization, 26% better than FineZip/LLMZip on the same model
SmolLM2-135M by
- 1 reply
liked a model 18 days ago
liked a model 23 days ago
liked a model about 1 month ago
upvoted a collection about 1 month ago
liked 2 models about 1 month ago
upvoted 2 articles about 1 month ago
Article
LightOnOCR-2-1B: a lightweight high-performance end-to-end OCR model family
ā¢
87
Article
We Got Claude to Build CUDA Kernels and teach open models!
- +2
ā¢
149
liked a model about 1 month ago
liked 3 Spaces about 2 months ago
liked 2 models about 2 months ago
