Quiet voice typing models

Speech models used by the Quiet messenger's voice typing (Android app). They run entirely on the phone: speech is turned into text on the device and is never sent anywhere. The app downloads one file, only when the user switches voice typing on, and uses it only if its SHA-256 matches the value built into the app.

These are 8-bit (ggml Q8_0) conversions for whisper.cpp of public fine-tuned Whisper models. No retraining; only the file format and number precision were changed.

File Based on Licence SHA-256
telugu-small-q8.bin vasista22/whisper-telugu-small Apache-2.0 355cef20a0d433ca6ffae35d414c817e0aeecfce21b934d68203efee1e72dcba
telugu-base-q8.bin vasista22/whisper-telugu-base Apache-2.0 818a290374cce182a5156c6b8e2c5b823109bd4ca97afbdc9dfd8eb788f23e39
hindi-base-q8.bin collabora/whisper-base-hindi CC BY 4.0 9b33b6720f0a5ee7e979ecc63765e995ef932d9c24d259639aca32b7f005d955

Credits: the Telugu models by vasista22; the Hindi model by Collabora; all based on OpenAI Whisper (MIT). Each file keeps its source model's licence (the Hindi file is CC BY 4.0, the Telugu files Apache-2.0). The conversion tool is in Quiet's source, tools/voice-models/.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support