Offline transcription assets The ONNX conversion is Xenova/whisper-base, revision 64da57285918e20ea79ea5c88eed7197933abaa8 (Apache-2.0 per its model card). https://huggingface.co/Xenova/whisper-base/tree/64da57285918e20ea79ea5c88eed7197933abaa8 Its base model is OpenAI Whisper (MIT): https://github.com/openai/whisper Model weights are redistributed unchanged. Chunking changes transport only. The accompanying model card and Apache-2.0 / Whisper MIT texts are supplied here. Transformers.js is Apache-2.0. Its bundled dependency notices are in /dependency-licenses.md. ONNX Runtime is MIT; its accompanying MIT notice is ONNX-Runtime-MIT.txt. The selected runtime is distributed unchanged. On-device language inference uses Qwen/Qwen2.5-1.5B-Instruct-GGUF, revision 91cad51170dc346986eccefdc2dd33a9da36ead9 (Apache-2.0 per its model card). https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct-GGUF The GGUF weights are redistributed unchanged and staged at build time. They run through @wllama/wllama (MIT, Copyright (c) 2024 Xuan Son NGUYEN, https://github.com/ngxson/wllama); the WASM runtime is distributed unchanged. See Qwen-LLM-model-card.md. Noto Sans JP and its SIL Open Font License are prepared together and included in exported editing packages. Browser rendering / codec notices, corresponding source, modifications, rebuild instructions, and Remotion usage-reporting details: /assets/browser-export/ The SDK attempts usage reporting when exporting, including while disconnected. No video, audio, subtitle text or editing data is included in that usage payload. Network failures are reported through the SDK logger; they do not switch to a cloud rendering route. This is not a claim of vendor approval of offline usage. Sources and license metadata checked 2026-11-03. User/native quality and browser acceptance are tracked in issues #7387 and #7384; general activation remains off.