Implement sharpening for thumbnails & previews about peertube HOT 3 OPEN

Hi,

Thanks for this interesting issue. Do you have resources on what value would be best (depending on image size and ffmpeg CPU) for -q:v or unsharp?

FYI ffmpeg doesn't scale the generated image, we use jimp (and so jpeg-js) for that purpose.

from peertube.

From my testing, -q:v 5 or 4 are a good setting for both thumbnails and the previews. In theory, the bigger the image the less noticeable the artifacts are and thus you can use higher values, but since the previews get displayed at higher than their native resolution of 850x480, keeping artifacts to a minimum is probably a good idea.

In terms of CPU usage, the impact is so minuscule that I'd have to process a large batch of images on my 15W dual-core from 2015 to even measure it. It's basically an instant operation.

from peertube.