README.md

---
license: mit
base_model:
  - openai/whisper-base
  - openai/whisper-small
  - openai/whisper-large-v3-turbo
library_name: whisperkit
pipeline_tag: automatic-speech-recognition
tags:
  - audio
  - automatic-speech-recognition
  - whisper
  - coreml
---

# whisperkit-coreml

OpenAI Whisper in CoreML format, published by [Diction Labs](https://diction.one) for on-device
speech-to-text on Apple hardware. Everything here is our own build, made with
[whisperkittools](https://github.com/argmaxinc/whisperkittools) (MIT) from OpenAI's
MIT-licensed Whisper checkpoints.

## Use with WhisperKit

```swift
let config = WhisperKitConfig(
    model: "openai_whisper-base",
    modelRepo: "DictionLabs/whisperkit-coreml"
)
let pipe = try await WhisperKit(config)
```

Runs on [argmax-oss-swift](https://github.com/argmaxinc/argmax-oss-swift) (MIT). Tokenizers are
resolved separately from the matching `openai/whisper-*` repo, so a first run still needs network
access even with the weights already on disk.

## Variants

| Folder | Base model | Size |
|---|---|---|
| `openai_whisper-base` | [`openai/whisper-base`](https://huggingface.co/openai/whisper-base) | full precision |
| `openai_whisper-small` | [`openai/whisper-small`](https://huggingface.co/openai/whisper-small) | full precision |
| `openai_whisper-large-v3-turbo` | [`openai/whisper-large-v3-turbo`](https://huggingface.co/openai/whisper-large-v3-turbo) | full precision, 1.63 GB |
| `dictionlabs_whisper-large-v3-turbo-q6q8` | [`openai/whisper-large-v3-turbo`](https://huggingface.co/openai/whisper-large-v3-turbo) | compressed, 703 MB |

`dictionlabs_whisper-large-v3-turbo-q6q8` is a compressed version of the turbo model above,
same weights, reduced precision, about half the size. Verified against the full precision
build with real transcription tests, not just internal accuracy checks:

| Language | Full precision (WER/CER) | Compressed (WER/CER) |
|---|---|---|
| German | 2.8% | 3.4% |
| English | 15.5% | 15.5% |
| French | 7.8% | 8.8% |
| Japanese | 2.6% | 2.6% |
| Korean | 18.1% | 22.5% |
| Chinese | 9.0% | 9.0% |

## Licence

MIT, same as upstream Whisper.