@aeva I’m not super sure how much AI image generation models struggle with noise, but modern video & image compression algorithms struggle quite a bit which may be part of why the noiseless look has become so desirable. I’m certainly in the camp that thinks noise can be pretty, and in modern films, it is often added in post.
Subjectively, I think *chroma* noise is what makes noise look bad. Try denoising only the chroma planes of your render with some wavelet denoiser and see if you agree.
If you're interested in video encoding at all, and you'd have a good time reading a video compression performance deep-dive with relevant metrics and a popular high efficiency video encoder, see my friend Trix's latest blog post on the Codec Wiki about SVT-AV1 v2.2's performance: https://wiki.x266.mov/blog/svt-av1-third-deep-dive
Just wanted to give a quick shout-out to my friend & fellow SVT-AV1-PSY developer Julio - check out his impressive portfolio in his new site! https://juliobbv.com/
@colin There are number of reasons why SVT-AV1 is faster than aomenc & rav1e - it is hard to compare to x265, since that encoder is far more mature & exists in service of a much more mature codec. I'm not even sure I'd say SVT-AV1 is faster in the first place...
vt-enc is a bash script that simplifies the process of encoding videos with FFmpeg using Apple's VideoToolbox framework on macOS. It provides an easy-to-use command-line interface for encoding videos with various options, including codec selection, quality settings, and scaling.
HEVC is also now supported on Discord clients with hardware accelerated HEVC decode, so your output will be Discord compatible.
My friend wrote a comprehensive guide to getting started with AV1 encoding a while ago called AV1 Encoding for Dummies. I've just written & published the modernized AV1 for Dummies guide on the Codec Wiki - you can check it out here: https://wiki.x266.mov/blog/av1-for-dummies
@renchap Luckily, it seems the image generation stuff & ChatGPT integration are both relatively tucked away, and it isn't necessarily shoved down your throat (unlike on Microsoft platforms...)
Explore more insightful benchmarking of the latest SVT-AV1 2.1.0 encoder and how it fares when encoding animation. The author, Trix, spared no expense in creating an incredible number of detailed graphs to support their conclusions about which settings are the best and why. Give it a read!
@jkoan@hammy@nate@radix Hi, we're still unfortunately dealing with registration. The IRS has requested more information twice now, and it usually takes around 4 months to get a response.
Really into codecs. Co-develops Aviator on Flathub. Immutable filesystem fan, JXL evangelist. CEO of The Radix Project. Studying at WPI. I'm searchable.#opensource #linux #foss