README.md

---
license: apache-2.0
library_name: vokra
tags:
  - vokra
  - gguf
  - speech-recognition
---

# whisper-medium (Vokra GGUF)

Speech recognition model, converted to the Vokra GGUF format for [Vokra](https://github.com/ayutaz/vokra), a zero-dependency speech-AI inference runtime.

**This is a conversion, not a new model.** The weights are the upstream ones; Vokra re-packages them so its runtime can memory-map them directly. Credit for the model belongs upstream — see *Source* below.

## Files

| File | Size | SHA-256 |
|---|---|---|
| `whisper-medium.gguf` | 2914.4 MB | `5674fc56480fb45caa90a0e68fd25168de15200c58c5af20e550dfbdc3d6ecf3` |

## Usage

```bash
# Download (any HTTP client works — the file is a plain GGUF)
curl -L -o whisper-medium.gguf \
  https://huggingface.co/vokra/whisper-medium/resolve/main/whisper-medium.gguf
```

```bash
vokra-cli run --model whisper-medium.gguf --input speech.wav

# Beam search instead of greedy
vokra-cli run --model whisper-medium.gguf --input speech.wav --beam-size 5
```

Prints the transcript. Input must be mono 16 kHz WAV.

## Provenance

| Field | Value |
|---|---|
| Architecture | `whisper` |
| Tensors | 947 |
| Upstream source | upstream distribution source (licence apache-2.0 per source) |
| Upstream licence | `apache-2.0` |
| Licence class | `permissive` |
| Registry model id | `whisper` |
| Vokra GGUF schema | 1 |
| Converted by | vokra-core 0.1.0-alpha.0 |

Every row above is read out of this file's own `vokra.*` metadata, so the card cannot claim something the artifact does not carry.

## Licence

The weights are distributed under **apache-2.0**, unchanged from upstream. Conversion does not alter the licence, and your obligations run to the upstream author.

## Verifying this file

```bash
shasum -a 256 whisper-medium.gguf
# expect: 5674fc56480fb45caa90a0e68fd25168de15200c58c5af20e550dfbdc3d6ecf3
```