Brushi includes open-source software and third-party models used for on-device speech and vision-language coaching. All coach inference and bundled voice synthesis run locally on your iPhone or iPad; Brushi does not send your child’s drawings to Senscilab, Google, or Apple cloud services for coaching.
Below are copyright and license notices for components shipped in or used by the app. See our Privacy Policy for how data is handled.
Brushi’s spoken coloring coach analyzes the canvas using LFM2.5-VL-1.6B model weights delivered through Liquid AI’s Leap SDK. Model files may be downloaded once from Liquid’s registry and cached on the device. Inference runs entirely on-device.
Copyright © Liquid AI, Inc. The LFM2.5-VL model is provided under the LFM Open License v1.0. Redistribution and use are subject to that license, including attribution and commercial-use terms.
Leap SDK runs GGUF model inference using a llama.cpp-compatible on-device stack (for example, llama.cpp/image-to-text inference types). Brushi does not call cloud Llama or Meta APIs.
Copyright © The llama.cpp authors. Licensed under the MIT License.
Brushi does not bundle Apple FastVLM model weights. We acknowledge Apple Machine Learning Research’s work on efficient on-device vision-language models, including FastVLM, which informs the broader ecosystem of edge VLMs that products such as Brushi build upon.
Apple Machine Learning Research models (including FastVLM) are subject to the Apple Machine Learning Research Model License Agreement. Where applicable, attribution: “Apple Machine Learning Research Model is licensed under the Apple Machine Learning Research Model License Agreement.” Copyright © Apple Inc. All rights reserved.
Brushi does not include Google Gemini models and does not send data to Google Gemini or other Google cloud AI APIs. If you use other Senscilab products that integrate Google services, those products are governed by Google’s applicable terms and privacy policies separately from Brushi.
English coach speech is synthesized offline using Sherpa-ONNX with bundled voice models. Parents may also choose Device voice, which uses Apple’s on-device text-to-speech (see Section 3).
Offline text-to-speech inference runtime from the sherpa-onnx project (k2-fsa).
Copyright © The sherpa-onnx authors. Licensed under the Apache License 2.0.
The Alba coach voice uses the Piper en_GB/alba ONNX voice from the rhasspy/piper-voices collection, synthesized through Sherpa-ONNX.
Piper is developed by Rhasspy. Voice weights are licensed under the MIT License.
Optional English coach voices Sky, Bella, Emma, and Sarah use the Kokoro en-v0.19 ONNX model (Kokoro-82M) via Sherpa-ONNX.
Copyright © hexgrad and contributors. Kokoro model weights are licensed under the Apache License 2.0.
Piper and Kokoro voice bundles include espeak-ng data for phonemization.
Copyright © espeak-ng authors. Licensed under the GNU General Public License v3.0.
When you select Device voice, or when the coach language is not English, Brushi uses Apple AVSpeechSynthesizer with voices installed on your device (for example, Martha or Kate for British English, or system voices for Arabic, Spanish, French, German, Japanese, Korean, and Chinese).
Apple voice synthesis is provided by Apple Inc. and is subject to your device software license. Brushi does not upload speech or drawings to Apple for this feature.
Brushi artwork, mascot characters, app design, and original coloring templates are © 2026 Sense Science Lab Ltd. All rights reserved.
For licensing or attribution questions, contact hello@senscilab.com.