rav1e VS SVT-AV1

There should be some major rav1e improvements in the not-so-distant future that should benefit high fidelity encoding very nicely (https://github.com/xiph/rav1e/pull/2921). And I hope someday someone gets 10/12 bit tune butteraugli working in aomenc, should provide a massive efficiency boost in pretty much all scenarios.

There's 3 that I know of,

libaom https://aomedia.googlesource.com/aom, the reference implementation.

https://gitlab.com/AOMediaCodec/SVT-AV1 - from the Alliance for Open Media, which Intel is deeply involved with and is spending a lot of time optimising.

https://github.com/xiph/rav1e - written in rust, also subject to a lot of optimising. It's not complete coverage of all features, intended to be used for places where libaom is too slow.

AV1 currently gets used by Netflix, YouTube (right click on any big popular video "stats for nerds) and so on. It's open source and currently has 3 different encoders. AOM as it's reference encoder available here: https://aomedia.googlesource.com/ as well as SVT-AV1 as the production encoder (like x265 for H265) available here: https://gitlab.com/AOMediaCodec/SVT-AV1 and Rav1e as community project which I personally don't have a use for at all: https://github.com/xiph/rav1e

Not AFAIK. The only relevant effort I'm aware of is rav1e, and it has more assembly in it than Rust.

No. There are hardware encoders. For software encoding while AV1's reference encoder is very slow; both rav1e and SVT-AV1 are usable on a higher end PC built in the last few years.

I'm going to promote rav1e, an AV1 encoder written in Rust+ASM. Unlike the reference encoder, rav1e seeks to focus on psychovisual quality rather than pure metrics, and so to many people it is a promising alternative. However, it has some catching up to do in the speed department.

Does rav1e have good documentation? From what I've seen, the best AV1 documentation by a good measure is that of SVT-AV1. Scroll down and you will see it. I would agree that they need a document on how to use it as a library. Maybe you can submit a feature request on gitlab.

I'm trying to follow this guide but -svtav1-params feature doesn't seem to be available. Any idea why?

Github Gitlab

If it shows up in 1.1 but not 0.87, I'd try a few intermediate versions (1.0, 0.9) and see where the artifacts are introduced. Then file a bug report. Give them a very short clip where it reliably shows up and the version in which you first notice it and I bet they will find and fix it right away. That will help all of us.

All three of the AV1 encoders work. I use SVTAV1, and I use it with ffmpeg. You just have to download a recent static master build until ffmpeg starts releasing official versions with post 1.0 SVTAV1. SVTAV1 can be very fast or it can be pretty thorough/efficient, and it makes good use of many-core computers. Some people feel that it ends up smoothing out very fine details more than other encoders, I think as part of its noise-removal algorithm. Its results seem fine to me. It is well documented (scroll down to the README) and includes ffmpeg examples and some of the tips/tricks you request. SVT-AV1 is the "official" production encoder for the AOM software implementation group.

AOMediaCodec's GitLab Link