---
license: mit
pipeline_tag: automatic-speech-recognition
---
# Earshot models

Distribution mirror maintained by StrumWorks Media for Earshot's optional on-device transcription model. This is a mirror of existing weights, not a model trained by StrumWorks Media.

Turbo: `ggml-large-v3-turbo-q5_0.bin`, 574,041,195 bytes.
SHA-256: `394221709cd5ad1f40c46e6031ca61bce88931e6e088c188294c6d5a55ffa7e2`.

Source: [ggerganov/whisper.cpp](https://huggingface.co/ggerganov/whisper.cpp/tree/5359861c739e955e79d9a303bcbc70fb988958b1), revision `5359861c739e955e79d9a303bcbc70fb988958b1`.
Based on OpenAI Whisper large-v3-turbo; quantized for whisper.cpp. Preserve the accompanying MIT licenses and original notices.

Earshot downloads the model file and performs transcription locally. Model hosting does not receive recordings or transcripts from Earshot. Hugging Face handles download requests under its own privacy policy.

The Small model remains bundled with Earshot. No user audio or transcripts belong in this repository.