AI Model Licenses

Last updated: 8 November 2026

About this page

Some FastBar tools use AI models made by others. They run only on your computer: your audio and video are never sent anywhere. Each model is downloaded from our server the first time you use its tool, together with its license files, which stay in %LOCALAPPDATA%\FastEdit Hub\tools. We may replace a model with a newer or better one; this page always lists the ones in use.

Transcript: Orukeet

Orukeet by Oruk AI, a fine-tuned version of NVIDIA Parakeet TDT 0.6B v3, speech recognition in 25 languages. Weights: CC BY-SA 4.0 (Oruk AI), based on NVIDIA Parakeet TDT 0.6B v3 (CC BY 4.0, NVIDIA). Audio preprocessor: Ivan Stupakov, parakeet-tdt-0.6b-v3-onnx (CC BY 4.0). Decoder graph composition: Derek Zeng (MIT). Source: huggingface.co/oruk/orukeet. We redistribute the files unchanged, in their combined INT8 ONNX form.

Beat Detect: Beat This!

Beat This! by Francesco Foscarin, Jan Schlüter and Gerhard Widmer (Institute of Computational Perception, JKU Linz), beat and downbeat tracking. Code and weights: MIT, github.com/CPJKU/beat_this. We converted the final0 weights to ONNX without changing them. Foscarin et al., "Beat this! Accurate beat tracking without DBN postprocessing", ISMIR 2024.

Other programs

These are not AI models, but the tools download them too, the first time they are needed, from their official GitHub releases (not from us), into the same folder. ffmpeg (Transcript, Silent Cut, Download from URL): the FFmpeg project, a build by the yt-dlp team, GPL, github.com/yt-dlp/FFmpeg-Builds. yt-dlp (Download from URL): Unlicense, github.com/yt-dlp/yt-dlp. deno (Download from URL, helps yt-dlp read some sites): MIT, github.com/denoland/deno.

Your content

A model's license covers the model, not what you run through it. Transcribing a recording or finding its beats does not give you any rights to it: make sure you may use the music and speech in your projects.