Licenses and Model Notices

Third-Party Technology

English Voice Coach uses open-source and third-party technology for app UI, local speech recognition, local language assistance, text recognition, and offline voice playback. The selected models and runtimes allow app use under their current source licenses, with notice and attribution obligations.

Current Notice Areas

The release preparation record currently tracks Flutter, Android, whisper.cpp, Whisper model files, Gemma 4 / LiteRT-LM, Sherpa-ONNX, Kokoro voice files, LiteRT, and Android ML Kit text recognition. The first release uses direct upstream downloads with pinned hashes rather than an owner-hosted model mirror.

Use Permission Summary

Gemma 4 E2B LiteRT-LM, Kokoro/Sherpa-ONNX, Sherpa-ONNX, and LiteRT/LiteRT-LM are treated as Apache 2.0 components. OpenAI Whisper and whisper.cpp are treated as MIT components. These licenses allow use in this app, including production/commercial use, when required notices are included.

Gemma 4 Review

English Voice Coach uses the Gemma 4 E2B instruction-tuned model in LiteRT-LM format: gemma-4-E2B-it.litertlm, from the Hugging Face repository litert-community/gemma-4-E2B-it-litert-lm, based on google/gemma-4-E2B-it by Google DeepMind.

The current public source labels the Gemma 4 model Apache 2.0, and Google's Gemma 4 model card/license page lists Apache 2.0. This allows use in English Voice Coach when Apache 2.0 notices are included. The expected SHA-256 for this app is 181938105e0eefd105961417e8da75903eacda102c4fce9ce90f50b97139a63c. The first release uses direct Hugging Face download with hash verification.

English Voice Coach is not affiliated with, sponsored by, or endorsed by Google, Google DeepMind, Hugging Face, or the LiteRT-LM project. Model outputs may be incomplete or incorrect and are provided as language-practice suggestions.

Speech Recognition

English Voice Coach uses converted Whisper model files from the whisper.cpp model repository for local speech recognition. Whisper and whisper.cpp are MIT-license components. The app keeps source records and pinned hashes for the tiny, base, and small English model files used by setup and settings.

Offline Voice

English Voice Coach can use Sherpa-ONNX and a Kokoro English voice archive for offline voice playback. These are treated as Apache License 2.0 components. Android TextToSpeech may be used as a system fallback when the offline voice model is not installed or cannot run.

App And Android Dependencies

The app uses Flutter, Dart packages, AndroidX, Android ML Kit text recognition, LiteRT/LiteRT-LM, Sherpa-ONNX Android packages, and other Android runtime dependencies. Their notices are included through Flutter's license surface where available and through this public notice page for model and runtime components.

License Sources

Source records for the current model stack: Gemma 4 model card, Gemma 4 Apache 2.0 license, Gemma 4 LiteRT-LM Hugging Face repository, Kokoro model card, OpenAI Whisper MIT license, and Sherpa-ONNX Apache 2.0 license.

License Text

Apache License 2.0 components are used under the Apache License, Version 2.0. MIT components are used under the MIT License. Copyright notices and source records are preserved in the app release record. If you need a copy of the license notices for a specific release, contact support@barbutsa.com.