Acknowledgements

Lavoi stands on other people’s work. Thank you to each of them.

Models Lavoi downloads

• Parakeet TDT 0.6B v3 by NVIDIA, under CC BY 4.0. Speech to text. Lavoi uses Parakeet Ultra, a post-training of this model by moondream, converted to Core ML by FluidInference.
• Silero VAD by Silero Team, under MIT. Finding where speech starts and stops. Converted to Core ML by FluidInference.
• Qwen3 0.6B and Qwen3 4B Instruct 2507 by Qwen Team, Alibaba Cloud, under Apache 2.0. The clean-up. Quantized to 4 bits for MLX by mlx-community.

Software Lavoi is built with

• FluidAudio by FluidInference, under Apache 2.0.
• MLX Swift and MLX Swift LM by Apple (ml-explore), under MIT.
• swift-transformers and swift-huggingface by Hugging Face, under Apache 2.0.

This site

• Satoshi by Indian Type Foundry (Fontshare), under Fontshare Free Font Licence.
• Inter by Rasmus Andersson, under SIL Open Font License 1.1.
• JetBrains Mono by JetBrains, under SIL Open Font License 1.1.
• Ghost Fibers (React Bits) by David Haz, under MIT + Commons Clause.
• Paper Shaders by Paper, under Apache 2.0.