Open-source licences
The components CaptionSmith uses, and their licences
CaptionSmith is built with the open-source software, speech models and fonts listed below. Each one is licensed under its own licence, named next to it. Every one of these licences allows use in a paid product. The full licence texts come with each component at the address shown.
Speech engines
Downloaded by your browser from Hugging Face the first time you use them, and saved on your device.
| Engine | Model | Licence |
|---|---|---|
| Fast (English) | onnx-community/whisper-base_timestamped, from OpenAI's whisper-base | Apache-2.0 |
| Best (every language) | onnx-community/whisper-large-v3-turbo_timestamped, from OpenAI's whisper-large-v3-turbo | MIT |
| Translate (English captions from other languages) | onnx-community/whisper-medium_timestamped, from OpenAI's whisper-medium | MIT (conversion), Apache-2.0 (original) |
| Hindi | yashwork-byte/indic-conformer-600m-int8-onnx, from AI4Bharat's indic-conformer-600m-multilingual | MIT |
Software in the app
| Component | Used for | Licence |
|---|---|---|
| Transformers.js (@huggingface/transformers) | Running the speech engines in the browser | Apache-2.0 |
| ONNX Runtime Web (onnxruntime-web, onnxruntime-common) | The speech processing runtime, downloaded from cdn.jsdelivr.net | MIT |
| @huggingface/tokenizers | Turning engine output into words | Apache-2.0 |
| @huggingface/jinja | Templates inside Transformers.js | MIT |
| protobufjs | Reading engine files | BSD-3-Clause |
| flatbuffers, long | Reading engine files | Apache-2.0 |
| guid-typescript | Part of ONNX Runtime Web | ISC |
| platform | Part of ONNX Runtime Web | MIT |
| mp4box.js | Reading MP4 and MOV files | BSD-3-Clause |
| mp4-muxer | Writing the exported MP4 | MIT |
| emojilib | Emoji suggestions | MIT |
Fonts
| Font | Used for | Licence |
|---|---|---|
| Inter, Montserrat, Poppins, Anton, Bebas Neue, Oswald, Mukta, Hind, Baloo 2, Kalam | Caption fonts | SIL Open Font License 1.1 |
| Noto Sans Devanagari | Hindi text | SIL Open Font License 1.1 |
| Noto Color Emoji | Colour emoji in captions | SIL Open Font License 1.1 |
| Instrument Serif, Schibsted Grotesk, Geist Mono | The website and the app | SIL Open Font License 1.1 |
Word lists
| Source | Used for | Licence |
|---|---|---|
| CMU Pronouncing Dictionary | How English words sound, for Hinglish captions | BSD-2-Clause |
| alt12dicts (Alan Beale), from SCOWL | Which words are common English words | Public domain |
| Common Voice sentences (Mozilla) | How often words are used | CC0 |
| AI4Bharat indic-conformer vocabulary and settings | The Hindi engine's word pieces, served from this site | MIT |
Questions about these licences: [email protected].