Legal / AI model notices
Third-Party AI Model Notices
Last reviewed: 2026-05-30
Sonavi uses or evaluates third-party model components including OpenAI Whisper, whisper.cpp-converted Whisper artifacts, NVIDIA Parakeet TDT artifacts converted for Sherpa-ONNX, and selected Google Gemma 4 sources. A listing on this page does not mean that a model is publicly downloadable, enabled by default, or eligible for automatic selection.
Component Notices
| Component | Model-weight license | Source | Provider terms | Platform/access terms | Conversion/runtime license | Sonavi availability | Modifications |
|---|---|---|---|---|---|---|---|
| OpenAI Whisper and whisper.cpp-converted artifacts | MIT | https://github.com/openai/whisper | OpenAI identifies Whisper code and model weights as MIT-licensed. | The cited upstream GitHub source is public. | whisper.cpp software: MIT | Selected converted artifacts are publicly distributed by Sonavi as listed below. | Selected model artifacts are converted to whisper.cpp-compatible GGML format. |
| NVIDIA Parakeet TDT 0.6B v3 and Sherpa-ONNX | CC BY 4.0 | https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3 | NVIDIA identifies the base model as CC BY 4.0. | The cited upstream Hugging Face model page is public. | Sherpa-ONNX conversion/runtime tooling: Apache-2.0 | The converted Sonavi artifact is publicly distributed as listed below. | Sonavi distributes a Sherpa-compatible converted and INT8-quantized ONNX archive derived from nvidia/parakeet-tdt-0.6b-v3; it is not the raw NVIDIA model. The Apache-2.0 tooling license does not replace the base-model CC BY 4.0 attribution and modification obligations. |
| Google Gemma 4 E2B IT | Apache-2.0 | https://huggingface.co/google/gemma-4-E2B-it/tree/70af34e20bd4b7a91f0de6b22675850c43922a03 | Google's Gemma 4 model card identifies Gemma 4 as Apache-2.0 licensed. | Upstream or Sonavi-controlled access requirements are separate from the model license. | Any applicable Sonavi packaging/runtime licenses remain separate. | Sonavi hosts the pinned artifact, but the application restricts it to manual selection and validation, requires a runtime probe, and excludes it from automatic selection. | Sonavi packaging or conversion changes must retain the Apache-2.0 license and applicable notices. |
| Google Gemma 4 12B IT QAT W4A16 compressed-tensors checkpoint | Apache-2.0 | https://huggingface.co/google/gemma-4-12B-it-qat-w4a16-ct/tree/9c79b5e652ae36f02bb07d3ca29124a9d1b009bd | Google's Gemma 4 model card identifies Gemma 4 as Apache-2.0 licensed. | Upstream visibility does not establish Sonavi distribution or authorize a status change. | Any applicable Sonavi packaging/runtime licenses remain separate. | Approved-conditionally for release preparation and validation only. This model is not publicly available from Sonavi. | QAT/compressed-tensors checkpoint. Any Sonavi packaging or conversion changes must retain the Apache-2.0 license and applicable notices. |
Production Model Notice Links
For each model artifact that Sonavi makes publicly downloadable, Sonavi publishes the corresponding provenance metadata and applicable license and notice files in that artifact's hosted directory. Availability and verification status may differ by model; a notice entry does not represent that every tracked or conditional model is publicly available.
| Model ID | Engine ID | Version | model.json | LICENSE.txt | NOTICE.txt |
|---|---|---|---|---|---|
| whisper-tiny-en | whisper-cpp | 1.0.0 | model.json | LICENSE.txt | NOTICE.txt |
| whisper-base-en | whisper-cpp | 1.0.0 | model.json | LICENSE.txt | NOTICE.txt |
| whisper-tiny-multilingual-q5_1 | whisper-cpp | 1.0.0 | model.json | LICENSE.txt | NOTICE.txt |
| whisper-base-multilingual-q5_1 | whisper-cpp | 1.0.0 | model.json | LICENSE.txt | NOTICE.txt |
| whisper-small-en | whisper-cpp | 1.0.0 | model.json | LICENSE.txt | NOTICE.txt |
| whisper-small-multilingual-q5_1 | whisper-cpp | 1.0.0 | model.json | LICENSE.txt | NOTICE.txt |
| whisper-medium-en | whisper-cpp | 1.0.0 | model.json | LICENSE.txt | NOTICE.txt |
| whisper-large-v3 | whisper-cpp | 1.0.0 | model.json | LICENSE.txt | NOTICE.txt |
| whisper-large-v3-turbo | whisper-cpp | 1.0.0 | model.json | LICENSE.txt | NOTICE.txt |
| nvidia-parakeet-tdt-v3 | sherpa-onnx | 1.0.0 | model.json | LICENSE.txt | NOTICE.txt |
No Endorsement
Provider names and marks, including OpenAI, NVIDIA, k2-fsa, and Google, are used only to identify source and provenance. No sponsorship, endorsement, or official status is claimed.