alt.freeweights.speech / VibeVoice-ASR-BitNet

From:

microsoft <huggingface.co/microsoft>Newsgroups: alt.freeweights.speechSubject:

VibeVoice-ASR-BitNet

Date: 2026-11-14Commit: 66e7802Signed: 2DC28CC7B26A283B ✓

Automatic speech recognition model with 323M parameters. Gguf and safetensors weights, 13.0 GB in all. Published by microsoft under MIT.

323M params

13.0GBVibeVoice-ASR-BitNetgguf safetensorspermissive

FactsParams323MSize13.0 GBFormatsgguf, safetensorsTaskautomatic speech recognitionLicenseMITFrom sourcepermissive

$ fw get microsoft/VibeVoice-ASR-BitNet
# manifest sha256 27eba5796bf5… · 13.0 GB · 14 files
$ fw verify ./VibeVoice-ASR-BitNet microsoft/VibeVoice-ASR-BitNet

From sourceHosted (signed)microsoft/VibeVoice-ASR-BitNetLast check: 2026-11-14 21:23 UTC

Run at home Get help

Get help: custom fine-tune or private deploy. We review your setup before we quote.

Works with

Run33 Fine-tune7 Convert9 Scan2

Runs with

Ranked by fit to this model (gguf, safetensors), then by GitHub stars.

BitNet tinygrad faster-whisper llamafile WhisperX FunASR NeMo Speech

Show all 33

Overview

Size: 13.0 GB
Weights: 703 MB (smallest of 2 GGUF variants)
Parameters: 323M
Formats: gguf, safetensors
Task: automatic speech recognition
License: MIT (permissive)
Files: 14, each with a sha256
Publisher: microsoft
Added: 2026-11-14
Section: alt.freeweights.speech

Lineage

Original release from the publisher.

Where these bytes come from

Publisher at the pinned commit→You

Sources are tried in order. A file counts only when its size and sha256 match the manifest. Sources are checked when you ask for them, not on a timer.

Sources

Badge	Host	Last check	Result
From source	huggingface.co	2026-11-14 21:23 UTC	passed

Model card

From the publisher at commit 66e7802. Shown as plain text with links; images open at the publisher.

VibeVoice-ASR-BitNet#

GitHub Technical Report License: MIT

VibeVoice-ASR-BitNet is a compressed variant of VibeVoice-ASR optimized for real-time inference on edge CPUs — no GPU required. Through heterogeneous quantization, the model is compressed from 4.62 GB to 1.58 GB while achieving 1.6–2.3× faster inference than Whisper.cpp with real-time capability (RTF < 1) on as few as 3 CPU threads.

➡️ Code: microsoft/VibeASR.cpp

➡️ Report: VibeVoice-ASR-BitNet Technical Report

➡️ Base Model: microsoft/VibeVoice-ASR

Image: image

🔥 Key Features#

⚡ Real-time on CPU — RTF < 1 with 3+ threads on commodity x86 (AVX2) and ARM (NEON) hardware
📦 Compact — 1.58 GB total (2.9× compression from FP16), fits in edge device memory
🌍 Multilingual — English, Chinese, French, Italian, Korean, Portuguese, Vietnamese, and more
🔧 Custom SIMD Kernels — Fused operators within the ggml framework for both ARM and x86 platforms

Quantization Strategy#

|:-:|:-:|:-:|:-:|:-:|

| VAE Tokenizer | 1.31 GB | 0.65 GB | I8\_S | 2.0× |

| LM Decoder | 3.32 GB | 0.92 GB | I2\_S + Q6\_K | 3.6× |

| Total | 4.62 GB | 1.58 GB | — | 2.9× |

Evaluation#

Inference Speed#

| Threads | 1 | 2 | 3 | 4 | 6 | 8 |

|:-:|:-:|:-:|:-:|:-:|:-:|:-:|

| RTF | 1.98 | 1.08 | 0.77 | 0.63 | 0.49 | 0.42 |

| vs. Whisper.cpp | 2.28× | 2.12× | 1.86× | 1.86× | 1.71× | 1.55× |

Benchmarked on AMD EPYC 7V13 (AVX2+FMA) with 20s audio. Bold = RTF < 1 (real-time).

Accuracy (WER%)#

|:-:|:-:|:-:|:-:|:-:|:-:|:-:|

| MLC-EN | 7.82 | 8.25 | 8.40 | 13.57 | 12.39 | 11.36 |

| MLC-FR | 16.03 | 17.41 | — | — | — | — |

| MLC-IT | 15.67 | 17.23 | — | — | — | — |

| MLC-KO | 9.83 | 11.15 | — | — | — | — |

| MLC-PT | 22.41 | 24.87 | — | — | — | — |

| MLC-VI | 20.15 | 22.38 | — | — | — | — |

| AISHELL4 | 19.83 | 27.45 | — | — | 22.52 | 20.41 |

| AMI-ihm | 17.42 | 21.36 | 21.92 | 27.07 | 30.81 | 32.07 |

| AMI-sdm | 24.18 | 25.87 | 26.33 | 36.92 | 48.11 | 40.17 |

| AliMeeting | 36.21 | 40.58 | — | — | 38.75 | 39.27 |

| Fleurs-en | 4.73 | 5.21 | 4.09 | 3.99 | 6.84 | 4.93 |

| Fleurs-zh | 7.92 | 8.35 | — | — | 5.56 | 7.00 |

| Libri-clean | 2.17 | 2.41 | 1.49 | 1.98 | 2.78 | 1.58 |

| Libri-other | 5.84 | 6.27 | 3.13 | 3.60 | 6.81 | 4.01 |

| VoxPopuli | 4.92 | 5.18 | 5.26 | 7.19 | 8.63 | 6.46 |

Model Files#

| File | Size | Description |

|:-:|:-:|:-:|

| vibeasr-vae-encoder-i8_s.gguf | 0.65 GB | VAE tokenizer, I8\_S quantized (ready to use) |

| vibeasr-lm-i2_s-embed-q6_k.gguf | 0.92 GB | LM decoder, I2\_S quantized (ready to use) |

| model-*.safetensors | 10.7 GB | Original SafeTensors (for conversion) |

License#

This project is licensed under the MIT License.

Contact#

This project was conducted by members of Microsoft Research. If you have suggestions, questions, or observe unexpected behavior, please contact us at VibeVoice@microsoft.com.

Files

14 files · 13.0 GBWeights 703 MB (smallest of 2 GGUF variants). Sizes and sha256 from the signed manifest.

Hide all 14 files with their sha256

File	Size	sha256
.gitattributes	306 B	3d17fc7e405d0039a2d600923e4a102b7f7fc6577df814d7c391f81d98c4c2ad
README.md	4.12 KB	ec5efef050b03bc294ba21db1913a4ebbb842f258c82fd5bcef762d0abe42aa3
config.json	3.53 KB	4873cb753c97f042886a93adfef478978662c21169755670de20f2ff73fa4ca1
figures/report_overview.png	223 KB	25ff476ff353744e5b52d26021cf50a12dd150d1513402f2b0b127355d0e384e
generation_config.json	138 B	8c970692323e3ea0e9b8b0a4dca79388d31226e41f83c9fd6014804280ebf6e8
model-00001-of-00003.safetensors	5.00 GB	58cb328634bb4b7e5afcc4f14c43261a0c636c9031b0b085bd3ef53c131aaf19
model-00002-of-00003.safetensors	4.96 GB	410349401256a9a995424408f8f77a21cad5e53183e07207820d7b25e84682df
model-00003-of-00003.safetensors	1.30 GB	49c2d7591f9dcbbd6fb37baa58379211f21df30dbd3c6ff39cb8c81473660cdb
model.safetensors.index.json	120 KB	89f02fb82c9228373f199303996451b3b20cd8207186c2975fded695fcdd456b
tokenizer.json	7.03 MB	f7c9b2dba4a296b1aa76c16a34b8225c0c118978400d4bb66bff0902d702f5b8
tokenizer_config.json	1.29 KB	80d64fccf954f022eb43b7425eb45a133cbb92c76a5f24d4cea6dbe3d45548d9
vibeasr-lm-i2_s-embed-q6_k.gguf	993 MB	fbe273d8dc2f2433bb25f849e19d77ea65aaa2188d12c20cee987ab6f321e002
vibeasr-vae-encoder-i8_s.gguf	703 MB	4941c82608c253ec066b5cc74d3dd11a5c8fef96cccbc5b87359ef0fe4338df6
vocab.json	2.78 MB	ca10d7e9fb3ed18575dd1e277a2579c16d108e32f27439684afa0e10b1440910

VersionsEach time the pinned upstream commit or file list changed in this repo. Newest first.

Upstream commit	Date	Contents
66e78021ab pinned now manifest updated, same commit	2026-11-14 21:23 UTC	14 files · 13.0 GB
66e78021ab	2026-11-14 21:23 UTC	14 files · 13.0 GB

Use it

$ fw get microsoft/VibeVoice-ASR-BitNet
# manifest sha256 27eba5796bf5… · 13.0 GB · 14 files
$ fw verify ./VibeVoice-ASR-BitNet microsoft/VibeVoice-ASR-BitNet

fw.pyChecks the signature and every sha256, tries each source in order, resumes.

$ python3 fw.py get https://freeweights.allternit.com/manifests/microsoft/VibeVoice-ASR-BitNet.json --dir ./VibeVoice-ASR-BitNet
$ python3 fw.py verify ./VibeVoice-ASR-BitNet ./microsoft__VibeVoice-ASR-BitNet.json

hfThe Hugging Face CLI with one setting. Same repo id.

$ HF_ENDPOINT=https://freeweights.allternit.com hf download microsoft/VibeVoice-ASR-BitNet --revision 66e78021ab8f5f06133d1ab421ba4d348bda97c9 --local-dir ./VibeVoice-ASR-BitNet

transformersSpeech recognition with the transformers pipeline.

import os
os.environ["HF_ENDPOINT"] = "https://freeweights.allternit.com"
from transformers import pipeline
asr = pipeline("automatic-speech-recognition", model="microsoft/VibeVoice-ASR-BitNet", revision="66e78021ab8f5f06133d1ab421ba4d348bda97c9")
print(asr("audio.wav")["text"])

llama.cppDownload one GGUF variant (vibeasr-vae-encoder-i8_s, 703 MB), then serve it.

$ HF_ENDPOINT=https://freeweights.allternit.com hf download microsoft/VibeVoice-ASR-BitNet --include "vibeasr-vae-encoder-i8_s*" --local-dir ./VibeVoice-ASR-BitNet
$ llama-server -m ./VibeVoice-ASR-BitNet/vibeasr-vae-encoder-i8_s.gguf

OllamaMake an Ollama model from the downloaded file.

$ printf 'FROM ./VibeVoice-ASR-BitNet/vibeasr-vae-encoder-i8_s.gguf\n' > Modelfile
$ ollama create vibevoice-asr-bitnet -f Modelfile

Run at home

License

Name: MIT License
Id: MIT
Class: Permissive
Redistribution: yes
Listing: Hosted (signed)
Decided: License as declared by the publisher. No person read it. 2026-11-14
Text in: README.md
Text sha256: ec5efef050b03bc294ba21db1913a4ebbb842f258c82fd5bcef762d0abe42aa3
Terms gate: No extra terms beyond the license.

Read the license text

Licenses are checked against the license table. A custom license passes only after a person reads it.

Provenance

From:

microsoft <huggingface.co/microsoft>Newsgroups: alt.freeweights.speechSubject: VibeVoice-ASR-BitNetDate: 2026-11-14Commit: 66e7802Signed: 2DC28CC7B26A283B ✓

Upstream: microsoft/VibeVoice-ASR-BitNet
Pinned commit: 66e78021ab8f5f06133d1ab421ba4d348bda97c9
Manifest: VibeVoice-ASR-BitNet.json · .minisig
Manifest sha256: 27eba5796bf56b0f5ddf3cc0187401b5cad4cb58658ed353e2d4c62d4f4b228a
Signature: minisign, Ed25519, key 2DC28CC7B26A283B
Created: 2026-11-14 21:23 UTC
Generator: freeweights-manifest/0.1.0
Takedowns: None.

How to verify

$ curl -O https://freeweights.allternit.com/manifests/microsoft/VibeVoice-ASR-BitNet.json
$ curl -O https://freeweights.allternit.com/manifests/microsoft/VibeVoice-ASR-BitNet.json.minisig
$ minisign -Vm VibeVoice-ASR-BitNet.json -P RWQ7KGqyx4zCLXtUFOqvfv8L3g+8fpKJ1FsuxPyOYMF+4KufbUxKq1Oq
$ shasum -a 256 <file>   # compare with the Files tab

A manifest whose signature fails must be rejected. A file whose size or sha256 does not match must be discarded. Verify a download

Model tree

No related models in this catalog.Nothing here lists this model as its parent, and it names no parent.

Get it

Download

$ HF_ENDPOINT=https://freeweights.allternit.com hf download microsoft/VibeVoice-ASR-BitNet

Open in Freeweights Studio

Pull, verify and run it in the Allternit app on Desktop, web or phone.

Signed manifest

minisign, key 2DC28CC7B26A283B. VibeVoice-ASR-BitNet.json. Provenance. Takedown history: none.

Source

huggingface.co/microsoft/VibeVoice-ASR-BitNet

Related models

All of alt.freeweights.speech →

From: MIT <huggingface.co/MIT>Subject: ast-finetuned-audioset-10-10-0.4593

From:

onnx-community <huggingface.co/onnx-community>Subject: ast-finetuned-audioset-10-10-0.4593-ONNX

From: Edge0 <huggingface.co/Edge0>Subject: Audio8-TTS-Preview-0.6b

From: Edge0 <huggingface.co/Edge0>Subject: Audio8-TTS-Realistic-Air

From: facebook <huggingface.co/facebook>Subject: audiobox-aesthetics

From:

mlx-community <huggingface.co/mlx-community>Subject: chatterbox-4bit

VibeVoice-ASR-BitNet

Works with

Runs with

Fine-tune with

Convert with

Scan with

Overview

Model card

VibeVoice-ASR-BitNet#

🔥 Key Features#

Quantization Strategy#

Evaluation#

Inference Speed#

Accuracy (WER%)#

Model Files#

License#

Contact#

Files

Use it

License

Provenance

How to verify

Model tree

Related models