vibevoice asr needs about 5.9 GB of memory at Q4_K (a 4.8 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 4 quants, from 0.0 GB to 16.7 GB (F16).
8.33B parameter speech‑to‑text model for 60‑minute audio
vibevoice asr is a unified speech‑to‑text model that accepts up to 60 minutes of audio in a single pass. It outputs structured JSON with speaker, timestamp, and content. The model is built on the CrispASR architecture and released under the MIT license.
Good at
- Single‑pass 60‑minute transcription
- Structured JSON with speaker and timestamps
- Custom hotword support
Limits
- No vision support
- Requires 24 kHz mono audio
- Large memory for full precision
Best for
- Transcribing long meetings
- Generating speaker‑aware subtitles
- Creating structured meeting notes
Who it’s for. Developers and researchers needing accurate, long‑form speech transcription with speaker diarization and timestamping. The model is suitable for applications that process continuous audio streams and require structured output.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.0 GB | 0.5 GB | Fits | Get file → | |
| Q4_K | 4.8 GB | 5.9 GB | Fits | Get file → | |
| Q8_0 | 9.4 GB | 11.1 GB | Too big | Get file → | |
| F16 | 16.7 GB | 19.2 GB | Too big | Get file → |
For 8 GB VRAM, use the Q4_K quantized file (4.81 GB, 5.9 GB memory at 8k context). For 16 GB VRAM, the Q8_0 file (9.45 GB, 11.1 GB memory) is appropriate. The F16 file requires 16.66 GB and 19.2 GB memory, suitable for high‑end GPUs.
What GPU do you need for vibevoice asr?
These figures add the file size, an fp16 KV cache at an 8K context window and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is the main file (0.0 GB file, about 0.5 GB in use).
- 6 GB VRAM: the best quant that fits is Q4_K (4.8 GB file, about 5.9 GB in use).
- 8 GB VRAM: the best quant that fits is Q4_K (4.8 GB file, about 5.9 GB in use).
- 12 GB VRAM: the best quant that fits is Q8_0 (9.4 GB file, about 11.1 GB in use).
- 16 GB VRAM: the best quant that fits is Q8_0 (9.4 GB file, about 11.1 GB in use).
- 24 GB VRAM: the best quant that fits is F16 (16.7 GB file, about 19.2 GB in use).
Which vibevoice asr quant should you choose?
Q4_K (4.8 GB) is the middle option in this repository. The usual sweet spot between file size and quality, and the quant most people start with.
If memory is tight, the smallest file is the first one in the table at 0.0 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about vibevoice asr
How much VRAM does vibevoice asr need?
About 5.9 GB at Q4_K with an 8K context window. That is the 4.8 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run vibevoice asr on an 8 GB GPU?
Yes, with Q4_K: the 4.8 GB file needs about 5.9 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of vibevoice asr GGUF?
It depends on the quant: file is 0.0 GB, Q4_K is 4.8 GB, Q8_0 is 9.4 GB, F16 is 16.7 GB.
Where can I download vibevoice asr GGUF?
The files are on Hugging Face in the cstr/vibevoice-asr-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does vibevoice asr use?
Hugging Face lists the license as mit. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run vibevoice asr on 8 GB VRAM?
Yes, the Q4_K quantized model is 4.81 GB and requires about 5.9 GB memory at 8k context.
Does the model support multiple languages?
Yes, it supports over 50 languages and handles code‑switching automatically.
What audio format is required?
Audio must be 24 kHz mono; pre‑convert if necessary.