| --- |
| license: other |
| license_name: mixed |
| language: |
| - da |
| - en |
| tags: |
| - whisper.cpp |
| - ggml |
| - punctuation |
| --- |
| |
| # Eckermann models |
|
|
| The model files that the [Eckermann](https://github.com/bhark/eckermann) dictation app downloads on first run. The app pins each file by commit and SHA-256. |
|
|
| | File | What it is | Source | License | |
| |---|---|---|---| |
| | `ggml-edda-v0.2-q8_0.bin` | Edda v0.2 Danish speech recognition, converted to GGML and quantized to q8_0 | [danish-foundation-models/edda-v0.2](https://huggingface.co/danish-foundation-models/edda-v0.2) at `698f8d0`, by the Alexandra Institute in the CoRal project. Based on [openai/whisper-large-v3-turbo](https://huggingface.co/openai/whisper-large-v3-turbo) (MIT) | Apache 2.0 | |
| | `ggml-silero-v6.2.0.bin` | Silero VAD v6.2.0, unchanged | [ggml-org/whisper-vad](https://huggingface.co/ggml-org/whisper-vad), from [snakers4/silero-vad](https://github.com/snakers4/silero-vad) | MIT | |
| | `punct-da/` | Danish punctuation and casing restoration, weights as f16 safetensors | [Alvenir/bert-punct-restoration-da](https://huggingface.co/Alvenir/bert-punct-restoration-da) at `922d302`. Based on [Maltehb/danish-bert-botxo](https://huggingface.co/Maltehb/danish-bert-botxo) by BotXO (CC BY 4.0) | Apache 2.0 | |
| | `ggml-large-v3-turbo-q8_0.bin` | Whisper large-v3-turbo for English, GGML q8_0, unchanged | [ggerganov/whisper.cpp](https://huggingface.co/ggerganov/whisper.cpp) at `5359861`, from [openai/whisper-large-v3-turbo](https://huggingface.co/openai/whisper-large-v3-turbo) | MIT | |
| | `punct-en/` | English punctuation and casing restoration, weights as f16 safetensors | [felflare/bert-restore-punctuation](https://huggingface.co/felflare/bert-restore-punctuation) at `954108a`. Based on [google-bert/bert-base-uncased](https://huggingface.co/google-bert/bert-base-uncased) (Apache 2.0) | MIT | |
|
|
| ## How the files were made |
|
|
| - Edda: `models/convert-h5-to-ggml.py` from [whisper.cpp](https://github.com/ggml-org/whisper.cpp) at `231638a`, then `whisper-quantize` with `q8_0`. |
| - punct-da and punct-en: [`bench/punct_model.py`](https://github.com/bhark/eckermann/blob/main/bench/punct_model.py) in the Eckermann repo. |
|
|
| The licenses of the original models apply to these copies. |
|
|
