Models / Microsoft/ VibeVoice ASR HF

VibeVoice ASR HF

Microsoft · released Mar 2, 2026 · microsoft/VibeVoice-ASR-HF

Speech to textTranscribes a recording into words

Input: audio. Output: text.InputOutput
Type
Open weightsMIT License
Languages
51
Size
8.3B

Context measured in tokens

Our take

Written Aug 2, 2026

VibeVoice ASR HF is an 8-billion-parameter speech-to-text model from Microsoft with a permissive MIT licence. It turns audio into text across 51 claimed languages and is extremely fast on benchmark hardware, though its accuracy varies sharply with recording quality.

Who should pick it

Pick this for open-source transcription with a permissive licence, or for high-throughput batch processing. Use it for clean read-aloud English audio, or multilingual deployment where breadth beats verified accuracy. Skip it if you need reliable transcription of meetings or accented speech, or if you require hosted inference.

The case for it

  • Extremely fast on benchmark hardware: one hour of audio processed in roughly sixteen seconds.
  • Permissive MIT licence allows commercial use, modification and redistribution.
  • Strong on clean read-aloud English speech, with a word error rate well below its own overall average.
  • Broad language coverage claimed, including major world languages.

The case against it

  • Accuracy collapses on challenging audio: recorded meetings and accented speech both show error rates more than seven times worse than its clean-speech performance.
  • No hosted inference options in our data, so self-hosting is the only path.
00

How good is it?

TranscriptionTurning speech into text2.5 of 5Open ASR WER · 49th of 74

Words it gets right

93.7%

Misses roughly one word in 16, averaged over nine English test sets.

How fast it listens

221×47th of 62

an hour of audio in 16 seconds, on the board's own hardware. Your machine will differ.

Languages

51

Listed on the model card. The accuracy above is English only.

Where it struggles
Read aloudaudiobooks, clean recording1.6%
Podcasts and videoeveryday internet audio7.9%
Accented speechspeakers from many countries12.2%
Meetingsa room, several people, far microphone12%

Percentage of words wrong on each set, lower better. Bars are scaled to this model's own worst case, not to the board.

Which languages

English · Chinese · Spanish · Portuguese · German · Japanese · Korean · French · Russian · Indonesian · Swedish · Italian · Hebrew · Dutch · Polish · Norwegian · Turkish · Thai · Arabic · Hungarian · Catalan · Czech · Danish · Persian · Afrikaans · Hindi · Finnish · Estonian · aa · Greek · Romanian · Vietnamese · Bulgarian · Icelandic · Slovenian · Slovak · Lithuanian · Swahili · Ukrainian · kl · Latvian · Croatian · Nepali · Serbian · Filipino · Yiddish · Malay · Urdu · Mongolian · Armenian · Javanese

The figures above come from the Open ASR Leaderboard, an independent public test that runs every model on the same recordings. It is the only measurement of transcription quality we know of, so there are no other scores to show.

Also scored, on boards we give no mark for
European-accented speech 6th of 74Podcasts and video 23rd of 74Clean read speech 43rd of 74Recorded meetings 44th of 74Accented speech 52nd of 74Financial calls 58th of 74Harder read speech 65th of 74

These tests check whether a model follows instructions — a precondition for all the work above, but not a measure of how well that work is done, which is why they get no rating.

Every published score for this model9 scoresEvery figure we hold, from 9 boards, with who ran it and a link to the source — including the boards no rating above is built on.
221.2independentsource ↗
6.3independentsource ↗
12.2independentsource ↗
3.5independentsource ↗
12independentsource ↗
7.9independentsource ↗
1.6independentsource ↗
4.9independentsource ↗
01

Can you run it yourself?

Fits in memory
weights load entirely on the card
Spills to system RAM
some weights offload; much slower
Too large
will not load even with offload
est
size is calculated; the verdict could change by 10%

Comfortable fit

A card many people ownFits in memory

GeForce RTX 4090 · 24 GB

Weights at Q4_K_M5.2 / 24 GBest
Spare memory16.1 GB spare
Usable context262Kwhat the spare memory holds; no published limit on record
Decode speed160 tok/sest

Room to spare. 16.1 GB spare means a 10% error in the size would not change the answer.

One step upFits in memory

GeForce RTX 5090 · 32 GB

Weights at Q4_K_M5.2 / 32 GBest
Spare memory24.1 GB spare
Usable context262Kwhat the spare memory holds; no published limit on record
Decode speed285 tok/sest

Room to spare. 24.1 GB spare means a 10% error in the size would not change the answer.

On a MacFits in memory

Apple M1 (8-core GPU) · 16 GB

Weights at Q4_K_M5.2 / 16 GBest
Spare memory5.3 GB spare
Usable context66Kwhat the spare memory holds; no published limit on record
Decode speed10 tok/sest

Room to spare. 5.3 GB spare means a 10% error in the size would not change the answer.

Your hardware
Checking your profile…

Memory use by level

Against a 24 GB card.

Q4_K_M
recommended
5.2 GBest
Fits in memory
Q5_K_M
6.1 GBest
Fits in memory
Q8_0
9.2 GBest
Fits in memory

All 3 sizes here are calculated, not measured. We hold no measured file for this model, so each size comes from the parameter count and every verdict above inherits that uncertainty.

Check against your own machine → · All 71 devices, with every size →

02

When we formed this view

Dates behind this page

Aug 2, 2026BenchmarkScored 12.2 on Accented speechleaderboard
Aug 2, 2026BenchmarkScored 3.5 on Financial callsleaderboard
Aug 2, 2026BenchmarkScored 12 on Recorded meetingsleaderboard
Aug 2, 2026BenchmarkScored 2 on European-accented speechleaderboard
Aug 2, 2026BenchmarkScored 7.9 on Podcasts and videoleaderboard
Aug 2, 2026BenchmarkScored 1.6 on Clean read speechleaderboard
Aug 2, 2026BenchmarkScored 4.9 on Harder read speechleaderboard
Aug 1, 2026ListedListed on LLMapfirst indexed by our pipeline
Aug 1, 2026BenchmarkScored 221.2 on Open ASR RTFxleaderboard
Aug 1, 2026BenchmarkScored 6.3 on Open ASR WERleaderboard

Prices last checked 6h ago

What we do not know about this model yet

  • We hold no measured file for it, so all 3 sizes on this page are calculated from the parameter count.
  • Nothing we hold says whether an endpoint streams, so we do not show it either way.
03

Licence and identifiers

What the licence allowsMIT License, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.

Licence

MIT License

permissiveCommercial use allowed

Fully permissive: do anything with attribution. No patent grant, unlike Apache-2.0.

Identifiers

Architecture
Dense
Modality record
audio->text
Catalogue slug
microsoft-vibevoice-asr-hf

Machine-readable model card (omc.json) →

Something wrong on this page? Tell us