FFmpeg 7.0 Released With Native VVC Decoding & Multi-Threaded CLI

[brad0]

Nonsense. x264 created a bitstream that can be decoded by the same means for h264, yet circumventing the patents on h264. That was the whole point of x264 and x265. Or with the words of the videolan team:

Or how do you think it would be possible to publish an encoder under GPL that's violating existing patents? Maybe the MPEG LA even tried to sue them, yet obviously never had any success, otherwise it wouldn't have been part of ffmpeg for so many years.

Tell me you have zero clue without telling me you have zero clue. That was extremely dumb.

[dragorth]

I was originally referring to 1080p 30 fps non-HDR 8bit H.264 original / H.264 output. I observe around 5x real-time encoding performance. This is with 5950X (16 cores), RX 7900 XTX, 2 x 32 GB DDR4-3600 CL14, Samsung 980 Pro 2 TB (up to 5100 MB/s). The final video was around 800 MB and 100 minutes in size. So it took 20 minutes to render the video (around 670 kB/s). A 1,2 MB floppy RAID-0 system (8bit Intel 8271 Floppy disk controller, Intel 8257 DMA controller from the year 1977) can write around 250 kB/s. That was 47 years ago.

That is just copying memory, firstly. No processing on the content.

The GPUs are processing the content, and it is processing the expanded content, those huge numbers I gave before, to shrink it down to your 670kB/s file size. The actual work was an extremely complex algorithm that reduced your original 2,985,984,000 bits/s down to 670 kB/s. And your card can operate on a 4K video to do that.

Your expectations are warped.

[Artim]

Tell me you have zero clue without telling me you have zero clue. That was extremely dumb.

Says the one having zero clue lol. You can try to disprove me of course, show me the judges rulings that x264 underlies the same patents as h264 including all of its royalty payments. And you can bet your ass that if the MPEG LA saw even the slightest chance of success they would have tried that.

[OneTimeShot]

Well, in case you missed the previous mp3 fiasco, mp3 codecs became kosher for 100% libre distros only after the very last patent in the pool expired. Decoding of CBR (non join) stereo 128 kbps was free long before. I think all the technology was already invented by 1993 or 1994. So it still took something like 5 years longer for distros (maybe even longer for server and LTS distros) to accept the fact and start shipping the libraries.

Well with that logic, as long as they keep adding a patented feature to a protocol every now and then, it'll never be out of patent. Sure, if they add a feature to h264 today then you can't use it for 20 years, but if you don't use the post-2003 features of h264, you obviously don't need to license those patents...

[Artim]

Well with that logic, as long as they keep adding a patented feature to a protocol every now and then, it'll never be out of patent. Sure, if they add a feature to h264 today then you can't use it for 20 years, but if you don't use the post-2003 features of h264, you obviously don't need to license those patents...

It wastes more money than it could make. After all, it needs to be something that doesn't already exist, or it would be prior art and thus invalid.

[sferrell]

TLDR; Any word on Dolby AC-4 support?

[brad0]

Says the one having zero clue lol. You can try to disprove me of course, show me the judges rulings that x264 underlies the same patents as h264 including all of its royalty payments. And you can bet your ass that if the MPEG LA saw even the slightest chance of success they would have tried that.

That's not how anything works. You're proving my point.

[caligula]

Well with that logic, as long as they keep adding a patented feature to a protocol every now and then, it'll never be out of patent. Sure, if they add a feature to h264 today then you can't use it for 20 years, but if you don't use the post-2003 features of h264, you obviously don't need to license those patents...

That's how it works. Windows media video/audio is totally different these days. For AAC there's he-aac.

[pWe00Iri3e7Z9lHOX2Qx]

Says the one having zero clue lol. You can try to disprove me of course, show me the judges rulings that x264 underlies the same patents as h264 including all of its royalty payments. And you can bet your ass that if the MPEG LA saw even the slightest chance of success they would have tried that.

Genuine question. What do you think the patents (it isn't just a singular patent, there are tons of them) related to AVC / H.264 cover? You can't build anything that makes a compliant H.264 stream that isn't covered by at least one of the patents.

https://www.via-la.com/wp-content/uploads/Final-November-1-2023-AVC-Attachment-1.pdf

Why do you think Fedora and OpenSUSE need to work with Cisco (who is a contributor to the patent pool) to provide the users of their distro a working H.264 solution?

[dec05eba]

- HDR10 metadata passthrough when encoding with libx264, libx265, and libsvtav1

FFmpeg also supports HDR10 metadata for vaapi and nvenc but it seems like nvidia, intel and amd all have bugged/incomplete drivers that ignore the data (it seems to work for transcoding, so im not sure how thats done). I wish FFmpeg could patch the file with hdr metadata output when not supported by the drivers. There are tools to do this as a second pass (one made by google).