-
Notifications
You must be signed in to change notification settings - Fork 1.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
contrib: ui: add libaom. #5036
base: master
Are you sure you want to change the base?
contrib: ui: add libaom. #5036
Conversation
Widley inconsistent results :( 1080p, Preset 5 is sub 10fps Doesn't seem to get much faster than this. Preset 10 is errr. I need to add more precision to the FPS counter. <1fps Not much user for the typical end user but maybe worth leaving as a compile time flag? |
Slower presets are really slow for a negligible quality improvement, but the medium one are comparable to svt-av1 slower one, with less visual artefacts. |
Any thing is strange, i can not compile it.
|
The full log should contain some reference on what's gone wrong, the last lines are not useful. |
OK i upload the build.txt |
I guess you need more ram. |
You were right. Increased the memory from 2GB to 4GB and it works. |
aomenc is very very slow in Handbrake. Only have around 10 fps and 40% CPU utilization, while with notenoughencoder I have around 25-30 fps and 100% utilization. Something doesn't seem to be going right. |
Play with the tiles setting like I wrote above. |
I pay with tile-columns and tile-rows. But it's also slowly, very slowly. |
any news about libaom? Version 3.6.1 is out. |
any news about libaom? Version 3.7 is out. |
I'd also like to see this encoder implemented. In my tests, libaom-av1 produces about 15-20% lower bit rate than svtav1 for the same psnr, ssim, and vmaf scores. For me, the smaller size along with fewer artifacts are worth the extra encoding time, which can be almost completely eliminated by running multiple instances. I'm not suggesting Handbrake should manage parallelized encoding. Though, that would be nice. Even svtav1 and x265 make fuller use of cpu when I'm running two instances of them. Handbrake's customized nlmeans implementation having temporal awareness works great for removing flicker from old TV shows, with minimal quality loss. I was unable to get as good a result with ffmpeg's filters. Davinci Resolve worked well but that was a manual process requiring several steps to end up with an mkv containing av1 video and opus audio since it supports neither. So, I've been waiting to encode some old shows until Handbrake has libaom-av1. @Mister-XY The reason you get faster encodes with libaom-av1 (or aomenc) using other apps (NEAV1E, av1an, etc.) is they split the source into pieces and run multiple instances of the encoder in parallel, then merge the encoded pieces together. The slow speed and low cpu utilization you saw are normal for one instance due to this encoder's poor multi-threading. |
Preliminary work on integrating libaom, I am opening a PR in the meantime if someone wants to play around with it.
It mostly work, but abr gives random bitrates, maybe I am setting something wrong.
Tested on: