DIY AI Studio - Offline Transcription
Third-party notices (reviewed 25 October 2026)

The application downloads the following unchanged third-party components directly
from their published hosts. Audio is processed on the user's device. The model
is an ONNX quantisation of Whisper Tiny English, not a DIY AI trained model.

Whisper Tiny English model
Base: OpenAI Whisper, https://huggingface.co/openai/whisper-tiny.en
ONNX conversion: Xenova/whisper-tiny.en
Revision: 79fb389fc764e7c395bd330e9531d9d32ada7049
https://huggingface.co/Xenova/whisper-tiny.en/blob/79fb389fc764e7c395bd330e9531d9d32ada7049/README.md
Both model cards identify Apache License 2.0. See licenses/APACHE-2.0.txt.
The converted model is used unchanged. The asset manifest records sizes and
SHA256 hashes for the downloaded q8 encoder, decoder and configuration files.

Transformers.js 3.8.1
Copyright Hugging Face and contributors.
https://github.com/huggingface/transformers.js/tree/3.8.1
Apache License 2.0. See licenses/APACHE-2.0.txt.
The distributed JavaScript is downloaded unchanged from jsDelivr.
Its browser bundle includes @huggingface/jinja 0.5.3 (the version in the
Transformers.js 3.8.1 source lockfile), under the MIT License. See
licenses/JINJA-MIT.txt for its copyright and licence notice.

ONNX Runtime Web 1.22.0-dev.20250409-89f8206ba4
Copyright (c) Microsoft Corporation.
https://www.npmjs.com/package/onnxruntime-web/v/1.22.0-dev.20250409-89f8206ba4
MIT License. See licenses/ONNX-RUNTIME-MIT.txt.
The distributed runtime and WASM files are downloaded unchanged from jsDelivr.
ONNX Runtime's additional notices from source commit 89f8206ba4 are retained
in licenses/ONNX-THIRD-PARTY-NOTICES.txt. That upstream inventory spans ONNX
Runtime build configurations and does not imply every component is active here.

The licensors do not endorse DIY AI. Review recognition output: speech models
can omit or invent words and their performance varies by accent and recording.
No model files are included in the application's Git repository.