boram VS ccapture.js

I'm using the ffmpeg gui wrapper Boram to cut, crop and encode VP9 webms.

I also tried to find a better recording method. Since I’m now using canvas, there are some solutions to recording that seemed quite elegant. For example, paper.js uses requestAnimationFrame to achieve its smooth animation. There are libraries like ccapture.js that hooks into the various methods like requestAnimationFrame and setInterval, allowing it to render each frame separately.

I actually spent a fair bit of time on this and eventually got something to work. Someone named "spite" made a great library called CCapture for doing this, but it needs to be updated and repackaged for ES6. Right now it is painful to integrate into your project, but it could be super-easy. You can help to encourage someone to fix the issues here.

Check out CCapture.

The image capture step uses ccapture to ensure that the playback frame is converted to an image. All frame images to webm video.

ccapture.js basically let's you write out frames from a threejs camera, which you then stitch together with ffmpeg.

I use ccapture, using the png option, which will spit out a tar with a png of each frame, and then I stitch them with ffmpeg.

This video is 1800 frames (30 FPS * 60s)... a browser trying to save 1800 individual frames would lock up and die. Ccapture.js hooks into the browser's rendering code and captures the frames as a 'blob' - a binary object that is often used for streaming video - and then wraps them up into one single download.

p5.js runs the sketch, ccapture.js captures all of the frames (1800 in this video), and ffmpeg combines the frames into an MP4.

I use ccapture.js, which captures the frames as a blob in real-time then exports it as a .tar file full of jpgs when the sketch is finished (with 'finished' defined as reaching 1800 frames, in this case). This is a lot better than having the browser try to save 1800 jpgs the usual way.