Legal

Third-party licenses

ClipEmber transcribes speech and draws captions with open models and fonts. Here is who made them and under which license.

Speech recognition

  • NVIDIA Parakeet TDT 0.6B v3 © NVIDIA Corporation, licensed under CC BY 4.0 (model card), via the ONNX export istupakov/parakeet-tdt-0.6b-v3-onnx (CC BY 4.0). Changes: we quantized the weights of the encoder’s matrix multiplications to 8-bit integers with a 7-bit range; other layers are unchanged.
  • Whisper small by OpenAI, MIT License, run through faster-whisper (MIT) — language detection and languages outside Parakeet’s 25.
  • Silero VAD by Silero Team, MIT License — finding where speech starts and ends.
  • onnx-asr (MIT) and ONNX Runtime (MIT) — running the models.

Caption fonts

Captions use Anton (© The Anton Project Authors), Archivo (© The Archivo Project Authors), Bangers (© The Bangers Project Authors), Bebas Neue (© Dharma Type), Caveat (© The Caveat Project Authors), Inter (© The Inter Project Authors), Kanit (© The Kanit Project Authors), Montserrat (© The Montserrat Project Authors), Oswald (© The Oswald Project Authors), Poppins (© The Poppins Project Authors), Roboto (© The Roboto Project Authors) and Sora (© The Sora Project Authors). For scripts and emoji they don’t cover, captions fall back to Noto Sans, Noto Sans CJK, Noto Sans Arabic, Noto Sans Hebrew, Noto Sans Devanagari and Noto Emoji by Google. All are licensed under the SIL Open Font License 1.1.

Caption emoji

Colour emoji above captions are PNG images from Noto Color Emoji by Google, used unchanged except for resizing. The project licenses its images under the Apache License 2.0 and its fonts under the SIL Open Font License 1.1.