Producing Good Quality Videos - Senscape/Dagon GitHub Wiki

For anyone interested in getting good quality videos, it is highly recommended to working in lossless video sequences where the individual frames are output as opposed to using movie clips (although another alternative is with movies that use a lossless video codec). Lossless images tend to have larger file sizes, which probably puts a lot of people off using them but this larger file size is because they are pixel perfect.

You should work with these lossless file types until you're ready to make the very last version of your movie, which will be a lossy ogv format for small file sizes. The reason to leave the lossy format until the very end is because lossy compression introduces artefacts/noise into the image and propagates the more lossy compression that occurs. The net result tends to be quite blocky and blurry videos.

If you haven't heard of them, lossless file types are pixel-perfect representations of your image (e.g. png, tga, exr) and lossy file types (jpg, mpg, mp4, ogg/ogv, avi, mov) use compression techniques alter the actual image in order to reduce file sizes. Admittedly, some of those lossy movie types are just containers (mov, avi) that can use both lossless and lossy codecs but most tend to be lossy. As mentioned above, the more you re-encode with a lossy image type, the more the image quality degrades (see Wikipedia Lossy Compression for more info).

Probably the best format is OpenEXR since it retains a lot of additional image information that can be used in compositors but the files can be rather large. The much smaller PNG or TGA files would suffice perfectly well if you don't plan on going crazy into compositing.

For that matter, the same goes for your cube renders as well - always render to a lossless format (i.e. TGA/PNG/EXR and not JPG) because the final conversion to TEX format is lossy and you want to reduce this as much as possible.