Last updated: October 6, 2026
VoltType runs these models and engines on your own phone or PC. They were made by the people named below, who share them under the licences shown. We thank them. Each licence was checked at its source on the date above.
| Used for | Model and maker | Licence |
|---|---|---|
| Voice pack (offline dictation, 25 languages) | Parakeet TDT 0.6B v3 by NVIDIA, converted to ONNX by the sherpa-onnx project | CC-BY 4.0 |
| Notes writer (summaries, to-do lists, meeting notes) | Android: Qwen2.5-1.5B-Instruct. Windows: Qwen3-1.7B. Both by Alibaba Cloud (Qwen team) | Apache 2.0 |
| Speakers in meetings: finding the turns | pyannote segmentation 3.0 by pyannote | MIT |
| Speakers in meetings: recognising a voice | CAM++ from 3D-Speaker by Alibaba (ModelScope) | Apache 2.0 |
| Windows: GPU speed-up and language detector | Whisper models by OpenAI | MIT |
| Translation | OPUS-MT by Jörg Tiedemann and the Language Technology Research Group, University of Helsinki, converted to ONNX by VoltType | CC-BY 4.0 |
| Engine | Licence |
|---|---|
| sherpa-onnx (k2-fsa) | Apache 2.0 |
| ONNX Runtime (Microsoft) | MIT |
| MediaPipe (Google) — Android | Apache 2.0 |
| whisper.cpp and llama.cpp (ggml) — Windows | MIT |
| node-llama-cpp — Windows | MIT |
VoltType is not made, sponsored or endorsed by any of the makers above. The models were converted (to ONNX or a smaller number format) so they run on a phone or PC; that conversion is the only change. The apps also use standard open-source libraries from Android, Electron and Node.js under their own licences.