MotionShine01 / Look
Shot by shot
The picture is compared every half second to find where shots change, and a frame from each shot is described by Florence-2, an AI model that runs in your browser.
Free tool
Audio description, written and read for you.
Audio description tells people who cannot see the screen what is happening in it. Drop a video and an AI looks at each shot and writes a line about it; the lines are placed in the quiet moments between speech, and an AI voice reads them. Edit every line, then save the described MP4, the voice alone, or a WebVTT description track. It all runs in this tab, so your video is never uploaded.
Describing
Reading
Mix
The first time, the picture AI model (about 360 MB) is downloaded once. Lines go into quiet moments where they fit; check each one against the picture before you share the video.
Your video is looked at, described and voiced in your browser. Nothing is uploaded.
How it works
See it, say it, fit it in.
02 / Fit
Into the quiet
Each line goes into the first quiet moment in its shot that is long enough. Where none is, the video's sound is turned down while the line plays.
03 / Save
MP4, WAV or VTT
The described video, the descriptions alone, or a WebVTT track for players that read description tracks.
Questions
Good to know.
Is the AI description good enough to publish as it is?
Treat it as a first draft. The model names what it sees, people, objects, places, text on screen, but it does not know who someone is or why a moment matters. Read every line against the picture and rewrite what is wrong or missing.
What is a WebVTT description track?
A text file of timed lines, like subtitles, marked as descriptions. Web players that support it can read the lines with a screen reader or a voice, so viewers can turn description on and off.
What if my video has no pauses?
Then lines are read at the start of each shot and the video's own sound is turned down under them. You can change how far it goes down, or move lines by hand to better places.
Which voices and languages are there?
The descriptions are written in English. They can be read by any of the voices from our text to speech tool.
Is my video uploaded?
No. The model is downloaded once from Hugging Face and runs in this tab; your video stays on your computer.