The universal multimedia toolkit. A diverse, community-driven project; posts are by individuals.

Over 100 security-report emails received in September already - and we’re only halfway through the month. The chart is noisy, but the security workload has grown substantially. If you’re a large company relying on FFmpeg please donate ffmpeg.org/donations.html
3
15
224
11,820
FFmpeg is critical infrastructure for a lot of big companies. The security-report workload is rising fast. It would be nice if donations, sponsorships, and code contributions kept pace.
19
37
435
20,849
Ampere has clarified that the 2021/22 server was sent to VideoLAN. That server benefited the wider free software multimedia ecosystem. Ampere has also agreed to donate a server to SPI for use in FFmpeg CI. We thank Ampere for helping resolve this.
110
19
985
141,774
One piece of the record
3
2
139
25,834
JB: if it was, please arrange to have it shipped to Germany so it can be professionally hosted and put to use for FFmpeg.
2
1
122
17,059
A few "behind the scenes" shots from our @lexfridman podcast 1) Post interview photo 2) Post interview @TerryBlacks_BBQ 3) Fatigue after six hours of talking Thank you to all the people who took the time to write nice messages.
27
21
838
36,996
More hand written assembly from legendary "mkver", removing old mmx vc1 mc functions and adding more recent ssse3. Over 17 times faster than C.
13
55
997
35,210
Nice patch from Zuxy removing old MMX(EXT) code and writing a faster sse2 version. Over three times faster than C
2
20
301
27,901
Huge patchset from @TencentGlobal adding hand written ARM NEON assembly for VVC angular prediction. Over 20 times faster than C.
17
46
1,875
68,550
Every little helps with hand written NEON assembly, even for simple functions. Up to 2x faster than C
12
29
629
64,593
Replying to @claudiogonzalo

ALT Betty White Dab GIF

26
1,409
Today's hand written assembly patch from the legendary "mkver" upgrading old MMX code into new(ish) SSSE3: Over 14 times faster than C.
42
98
2,473
102,600
FFmpeg is mandating human verification of security issues as a result of a spike in AI generated false positives
45
78
2,001
141,278
Thank you to @ZTEPress for sending us hand written RISC-V assembly patches. Over 11 times faster than C!
9
57
1,495
117,184
Basically every other post on @ffmpeg is a screenshot of hand written asm benchmarks vs C. Often 10-20x faster.
1
34
1,993
Great pull request from @ruikai fixing security issues with patches. A sixteen year old showing security "researchers" how it's done. In FFmpeg, teenagers work on world leading C and assembly code.
7
26
987
52,000
Here's a great example of the compromises FFmpeg has to make because of the wide variations of ARM CPU cores out there
Replying to @FFmpeg
That's very counter intuitive, I'd assume you'd get more consistency on RISC than on CISC. But I guess Intel and AMD have an incentive to match the other's performance stronger than on ARM
8
11
285
41,353
Someone on the FFmpeg mailing list claims to have built a superoptimiser that beats hand written assembly. We'll find out soon enough
24
20
1,015
157,706
Patch by legendary "mkver", removing a few redundant functions in existing hand written assembly code. Speed improvement change from up to 16x faster than C to 18x faster than C. Assembly code is not necessarily ever finished.
9
19
762
37,687
Cool patchset to decode Broderbund Mohawk video files from 1990s games. FFmpeg aims to play every video file ever made.
8
10
234
11,800
More hand written assembly patches from legendary "mkver", removing old MMX code and adding SSE4 for niche audio codec SBC. Nearly ten times faster than C
14
19
441
22,442
Hybrid approach to crc32 with interleaved crc32 instruction and pmull hand written assembly patch. Over 67 times faster than C. Thanks to @corsix for the original implementation.
7
23
360
16,771
Hand written assembly patch by legendary "mkver" removing old MMX GMC code and replacing it with SSSE3 code Over ten times faster than C.
18
40
930
38,918
Patch for hand written @Arm assembly optimisations of VVC put_chroma_v Up to five times faster than C
4
15
318
12,064
Replying to @GoogleAtHome
17
1,774
RISC-V SAD hand optimised assembly language code patch in FFmpeg. Over ten times faster than C
5
7
219
13,040
Thanks to @ZTEPress for sending FFmpeg RISC-V patches!
3
128
7,908
Patch by legendary "mkver" adding median prediction hand optimised assembly code. Over 27 times faster than C.
8
21
555
21,048
Thanks to @mypopydev, we have a Chinese translation of our assembly lessons! github.com/FFmpeg/asm-lesson…
4
13
319
22,678
More hand optimised assembly language from legendary "mkver" Over 16 times faster than C
6
15
638
22,696
FFmpeg assembly lessons are now available in Turkish github.com/FFmpeg/asm-lesson…
13
72
1,130
40,569
OG FFmpeg developer @astrange_e on C++ developers
8
17
434
26,816
FFmpeg's Vulkan FFv1 encoder and decoder is being moved to SPIR-V. Here are the benchmark results on an RTX 5070
5
5
253
13,015
More ARM hand written assembly patches from @TencentGlobal today cleaning up existing assembly code. ARM assembly needs constant maintenance as algorithmic approaches may need changing as newer ARM CPUs are released. (Generally speaking x86 is not like that, code remains broadly the same speed forever)
6
26
848
46,079
We have commented on "value" before
6
3
143
6,839
"There's no value in FFmpeg on its own" according to this guy who doesn't understand the world value apart from its monetary meaning: piped.video/watch?v=77fH--tl…
25
20
534
62,416
Huge patchset from @TencentGlobal adding hand written assembly code for HEVC intra prediction Up to ten times faster than C! Companies which use FFmpeg should submit hand written assembly patches like Tencent
52
157
5,329
162,306
It's with much sadness that we announce the passing of our friend and colleague Ralph Giles, or rillian as he was known on IRC. Ralph began contributing to Xiph.org in 2000 and became a core Ghostscript developer in 2001¹ . Ralph made many contributions to the royalty-free media ecosystem, whether it was as a project lead on Theora or maintaining Xiph infrastructure that has been used across the industry by codec engineers and researchers². He was also the first to ship Rust code in Firefox³ during his time at Mozilla, which was a major milestone for both the language and Firefox itself. Ralph was a great contributor, a kind colleague and will be greatly missed. ¹ wizards-of-os.org/archiv/spr… ² media.xiph.org/ ³ medium.com/mozilla-tech/depl…
49
174
3,307
77,527
Google are sending patches using Codemender
9
6
426
53,857
We now have ten sponsors! github.com/ffmpeg (Didn't realise @DarthPutinKGB was such a fan!)
We have three Github sponsors. Thank you for your generosity!
11
15
540
26,057
We have three Github sponsors. Thank you for your generosity!
21
26
2,018
132,798
Great to see legendary "mkver" removing more legacy MMX code. No major speed improvements today but we love seeing old "emms" functions removed!
6
7
345
22,330
Patch for aarch64 NEON PMULL+EOR3 SIMD hand written assembly for CRC functions Up to 35x faster than C
8
12
309
112,275
Thanks to @bagder for covering the FFmpeg AI security situation @fosdem One Google employee is now sending patches and we are thankful piped.video/6wYSwZ20NJU?t=2790
1
1
75
9,665
Learn more about recent updates in FFmpeg from this weekend's @fosdem fosdem.org/2026/schedule/eve…
2
8
149
9,252
Many FFmpeg developers are at @fosdem today!
4
7
389
13,807
Hi @AMD. FFmpeg developer @quink_lamy is not happy with your AI slop patches. code.ffmpeg.org/FFmpeg/FFmpe…
43
130
3,319
205,126
The infamous example how not to do it, was NAB 2023 where a company (Videolabs) had a booth that appeared to represent FFmpeg possibly for commercial advantage without knowledge or consensus of the FFmpeg community. And not donating any Videolabs shares to FFmpeg/SPI.
1
37
5,927
Patch by legendary "mkver" to optimise HEVC dequant with hand written assembly. Over five times faster than C. Also a great example of how Clang is much slower than GCC in this case.
9
11
472
19,895
Indeed a significant portion of the "C code" in FFmpeg today is for various tables (e.g huffman codes) needed by codecs. Not really "C code" as such. The assembly portion of FFmpeg needs to accelerate.
Replying to @FFmpeg
Soon ffmpeg will become an assembly library instead of C. Performance is all that matters.
18
19
1,190
73,054
Patches with hand written ARM assembly for VVC put_luma functions. Up to 2.5x faster than C. The rate of assembly patches in FFmpeg is accelerating!
4
9
377
20,561
Patch for "MPEG Surround data, USAC variant" by Lynne
122
9,277
Goals
In the near future the entire ffmpeg's codebase is gonna be replaced with asm bruh😭
65
209
7,899
292,990
Hand written RISC-V assembly code written by @AlibabaGroup Cloud submitted to FFmpeg Up to 14 times faster than C. It's great to see so many corporate contributors of hand written assembly, a field historically dominated by volunteers!
75
365
6,818
275,981
Hand written assembly for H.264 16x16 intra prediction functions. Up to 3x faster than C. Even small improvements are valuable in FFmpeg.
13
40
594
37,062
Many HEVC arm64 hand written assembly optimisations today from @TencentGlobal Up to 12x faster than C
42
165
2,857
152,313