![]() |
||||||||||||||||||||||||||||||||||||||||||||||||
[HELP] AV1+NVENC Constant Quality - Printable Version +- Selur's Little Message Board (https://forum.selur.net) +-- Forum: Hybrid - Support (https://forum.selur.net/forum-1.html) +--- Forum: Problems & Questions (https://forum.selur.net/forum-3.html) +--- Thread: [HELP] AV1+NVENC Constant Quality (/thread-4194.html) |
||||||||||||||||||||||||||||||||||||||||||||||||
AV1+NVENC Constant Quality - tomek398 - 21.08.2025 I'm encoding 4K file from a GoPro (H.264) to AV1 using NVEnc Constant Quality. Encoder Preset: Quality Most recent NVidia Studio driver. No other changes to the input (same output resolution, frame rate etc....) I'm very surprised to see that the size of the output file doesn't really depend on the Quality factor. It looks like that: CQ16: 109,592MB CQ18: 109,592MB CQ20: 109,592MB CQ22: 109,592MB CQ24: 108,746MB CQ26: 108,735MB CQ28: 108,723MB Why is it like that? It is very different behaviour to the NVEnc H.265 where the output file size varies a lot when changing CQ factor RE: AV1+NVENC Constant Quality - Selur - 22.08.2025 It mainly depends on the amount of details and how they are preserved at different quantizers. Assuming crf scaled linear with file size is wrong. If there are no details, a lower quantizer will retain, lowering the quantizer / crf does not change the output. (Sometimes quantizer x-y can give the same output, but x-(0.1) or y+(0.1) can be a lot different.) I tested: a. an SD clip and got:
![]() With software encoders, you usually get way more diversity in sizes and detail retention, since they usually take more/better psycho visual effects into account. Cu Selur RE: AV1+NVENC Constant Quality - tomek398 - 22.08.2025 Thanks for replying. I wasn’t sure if this was an issue with the drivers or maybe some incorrect settings. My input is 4K, and I expected to see something similar to H.265, not a linear dependence. |