YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
InkVoice models
Every model the InkVoice app downloads. The folder layout is the app's own model folder: a file at
<folder>/<name> here is installed at <app data>/models/<folder>/<name>. Sizes and SHA-256 checksums are pinned in the
app (lib/services/shared_model_service.dart), which also pins this repo's revision.
| Folder | What | Source and licence |
|---|---|---|
litert/gemma-4-e2b/ |
Chat and vision: Gemma 4 E2B (.litertlm, 2.59 GB) |
Google, via litert-community. Apache-2.0 |
asr/indic-flex/ |
Speech recognition, Indian languages + English (int4 ONNX) | indic-transcribe-flex by Bodhan AI / AI4Bharat, Indic Open Model License v1.0 (built on NVIDIA Canary-1B-v2) |
asr/whisper-turbo/ |
Speech recognition, 99 languages with word times (q4 ONNX) | Whisper large-v3-turbo by OpenAI, MIT; ONNX build by onnx-community |
tts/omnivoice/ |
Voice (int4 ONNX). See tts/omnivoice/NOTICE.md |
OmniVoice by k2-fsa: weights CC-BY-NC 4.0 (non-commercial); Higgs Audio v2 tokenizer by Boson AI |
vad/ |
Silence detection: Silero VAD | Silero, MIT |
smart_search/ |
Smart search: EmbeddingGemma 2 text, image and audio encoders (q4 ONNX) | Google, Gemma terms; ONNX build by onnx-community |
diar/onnx/ |
Speaker detection: Nemotron-3 diarization (q4 ONNX) | NVIDIA, OpenMDW License 1.1; ONNX build by onnx-community |
Third-party files are unmodified copies of the upstream ONNX builds (same bytes, same checksums) so the app does not depend on other repositories staying available.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support