NbAiLab/nb-whisper-medium from the National Library of Norway (Nasjonalbiblioteket), converted to Core ML for WhisperKit so it runs on the Apple Neural Engine.
This is a format conversion, not a new model. The weights and behaviour are the National Library's. All credit for the model belongs to NbAiLab.
Used by the Mac, iPhone and iPad app Sagt for local Norwegian transcription («Norsk nøyaktig»).
whisperkit-generate-model --model-version NbAiLab/nb-whisper-medium)main (commit 0ed074d5985bd56ca4140159a9dbffbc3fb5117e)NbAiLab_nb-whisper-medium)config.json and generation_config.json are copied unchanged from the original model.Word error rate on real Norwegian speech, measured with the same clips, normalisation and code path on the Neural Engine. In our measurements NB-Whisper medium was as accurate as NB-Whisper large on every set, at half the size:
| Set | NB-Whisper medium (this) | NB-Whisper large |
|---|---|---|
| FLEURS nb (75 clips) | 6.5 % | 7.0 % |
| NST (75 clips) | 2.6 % | 2.2 % |
| NB Tale, spontaneous speech (75 clips) | 15.8 % | 17.1 % |
| NPSC, Storting (75 clips) | 14.6 % | 14.8 % |
| Storting, four ~10-minute meetings | 13.6 % | 13.9 % |
The original PyTorch model gives the same numbers within noise. NPSC and NB Tale have verbatim references, while NB-Whisper writes clean text (drops repetitions and false starts), which explains most of the error there.
let folder = try await WhisperKit.download(
variant: "NbAiLab_nb-whisper-medium",
from: "Selbyggen/nb-whisper-medium-coreml"
)
let kit = try await WhisperKit(WhisperKitConfig(modelFolder: folder.path))
let result = try await kit.transcribe(audioPath: "opptak.m4a",
decodeOptions: DecodingOptions(language: "no"))
The model uses the same tokenizer as openai/whisper-medium.
Sagt uses this model to transcribe recordings and meetings on Mac, iPhone and iPad. Everything runs on the device, and no audio leaves it.
Apache 2.0, the same as the original model.
NbAiLab_nb-whisper-medium/TextDecoder.mlmodelc/weights/weight.bin 913 MBNbAiLab_nb-whisper-medium/AudioEncoder.mlmodelc/weights/weight.bin 614 MBNbAiLab_nb-whisper-medium/MelSpectrogram.mlmodelc/weights/weight.bin 354.1 kBNbAiLab_nb-whisper-medium/TextDecoder.mlmodelc/coremldata.bin 0.7 kBNbAiLab_nb-whisper-medium/AudioEncoder.mlmodelc/coremldata.bin 0.4 kBNbAiLab_nb-whisper-medium/MelSpectrogram.mlmodelc/coremldata.bin 0.4 kBNbAiLab_nb-whisper-medium/AudioEncoder.mlmodelc/analytics/coremldata.bin 0.2 kBNbAiLab_nb-whisper-medium/MelSpectrogram.mlmodelc/analytics/coremldata.bin 0.2 kBNbAiLab_nb-whisper-medium/TextDecoder.mlmodelc/analytics/coremldata.bin 0.2 kB