Beta: these terms are being reviewed by counsel and may change.
Legal
Open-Source Notices
Last updated October 25, 2026
XCap is built on open-source AI models and software, and a few other freely licensed components. We're grateful to their authors. This page lists them with their licences, as those licences require. Copyright in each component belongs to its respective authors.
1. AI models
| Component | Used for | Licence |
|---|---|---|
| Wan 2.2 | Video generation (Fast and Quality clips) and talking presenters | Apache License 2.0 |
| umt5-xxl | Text encoder for Wan 2.2 (turns clip prompts into model input) | Apache License 2.0 |
| wav2vec2-large-xlsr-53-english | Audio encoder for talking presenters (follows the voiceover for lip-sync) | Apache License 2.0 |
| LTX-2.3 | "With audio" clips (video with its own sound) Free for commercial use by organisations under US$10M annual revenue; above that a commercial licence from Lightricks is required. The operator of this Service is responsible for staying within those terms. Its use restrictions apply to everyone who uses this Service: see "Rules from the AI models we use" in our Acceptable Use Policy. | LTX-2 Community License Agreement |
| Gemma 3 12B | Text encoder inside LTX-2.3 (turns video prompts into model input); used for nothing else Gemma is provided under and subject to the Gemma Terms of Use found at ai.google.dev/gemma/terms, including the Gemma Prohibited Use Policy (ai.google.dev/gemma/prohibited_use_policy). Its use restrictions apply to everyone who uses this Service: see "Rules from the AI models we use" in our Acceptable Use Policy. | Gemma Terms of Use |
| Qwen-Image 2512 | Image generation (pictures with readable text, start frames for Quality clips, stock presenter portraits) | Apache License 2.0 |
| FLUX.1 [schnell] | Image generation | Apache License 2.0 |
| Qwen3 | Text generation (scripts, captions, agents) | Apache License 2.0 |
| Qwen3-Embedding 0.6B | Text embeddings for XCap Score and for grouping trends | Apache License 2.0 |
| Qwen2.5-VL 7B | Checking generated images and clips (quality, nudity, gore, public figures and characters) and screening uploaded photos (no face recognition) | Apache License 2.0 |
| ACE-Step 1.5 | Music generation | MIT (Comfy-Org's repackaged weights are tagged Apache-2.0) |
| Kokoro-82M | Text-to-speech | Apache License 2.0 |
| Chatterbox | Text-to-speech and voice cloning | MIT License |
| faster-whisper | Speech-to-text (captions, transcripts) | MIT License |
| YuNet (OpenCV Zoo) | Face detection for framing and AI twins (no face recognition or biometric templates) | MIT License |
2. Software and fonts
| Component | Used for | Licence |
|---|---|---|
| FFmpeg | Audio and video processing Run as a separate, unmodified program: the ffmpeg-static and ffprobe-static builds, or the operating system's FFmpeg package on our Linux servers. These builds include GPL components such as x264. Source code is available from ffmpeg.org; we will provide the exact source of the builds we use on request. | GNU GPL (version 3 for the bundled static builds) |
| Ollama | Serving the text, vision and embedding models | MIT License |
| ComfyUI | Running the image, video, presenter and music models Runs as a separate, unmodified program on our own servers. Its source code is available at github.com/Comfy-Org/ComfyUI. | GNU GPL v3 |
| speaches | Speech-to-text server (runs faster-whisper) | MIT License |
| Kokoro-FastAPI | Text-to-speech server (runs Kokoro) | Apache License 2.0 |
| espeak-ng | Pronunciation inside Kokoro-FastAPI Used unmodified, inside the Kokoro-FastAPI server. Its source code is available at github.com/espeak-ng/espeak-ng. | GNU GPL v3 or later |
| chatterbox-tts-api | HTTP server for Chatterbox voices Runs unmodified (upstream commit a5f46612) as a separate program on our own servers. Its source code is available at github.com/travisvn/chatterbox-tts-api. | GNU AGPL v3 |
| onnxruntime | Running the YuNet face detector | MIT License |
| web-push | Encrypting and sending notifications to your devices Used unmodified. Its source code is available at github.com/web-push-libs/web-push. | Mozilla Public License 2.0 |
| resvg-js | Drawing thumbnails, carousels and app icons Used unmodified. Its source code is available at github.com/yisibl/resvg-js. | Mozilla Public License 2.0 |
| HyperFrames | Rendering motion graphics (animated captions, carousels and ad templates) to video | Apache License 2.0 |
| GSAP | Animation in motion-graphics templates and their previews Provided by GreenSock under its Standard "No Charge" License, which is not an open-source licence. Copyright GreenSock. | GSAP Standard "No Charge" License |
| Chrome headless shell | Capturing motion-graphics frames for HyperFrames Chromium's third-party licences are listed in the LICENSE.headless_shell file that ships with it. | BSD 3-Clause, plus Chromium's third-party notices |
| sharp | Image processing in HyperFrames | Apache License 2.0 |
| libvips | Image library used by sharp Used unmodified, as the prebuilt libraries that ship with sharp; they can be replaced. Its source code is available at github.com/libvips/libvips. | GNU LGPL v3 or later (prebuilt libraries) |
| Montserrat | Caption and interface font | SIL Open Font License 1.1 |
| Inter | Interface font | SIL Open Font License 1.1 |
| Noto fonts | Caption fonts for other scripts and emoji | SIL Open Font License 1.1 |
3. What this means for your content
These licences cover the models and software. Ownership of your output is set out in our Terms of Service. Two model licences also limit how the model and what it makes may be used: the LTX-2 Community License and the Gemma Terms of Use (with the Gemma Prohibited Use Policy). Their rules are part of our Acceptable Use Policy and apply to everyone who uses XCap; the full LTX-2 licence is linked in the table above.
4. Questions and source requests
For licence questions or source-code requests, contact XCap at [email protected].