← Home ← Back to /g/

Thread 107143844

25 posts 14 images /g/
Anonymous No.107143844 [Report] >>107145376 >>107146032 >>107148316 >>107151082 >>107152771
Does anyone else encode video with SVT-AV1 on their phone? I find the energy efficiency aspect fascinating.

#!/bin/bash
for f in *.mov; do
ffmpeg -i "$f" -c:v libsvtav1 \
-preset 8 -crf 40 \
-c:a libopus -b:a 128k "${f%.mov}.webm"
Anonymous No.107143875 [Report] >>107144020
its energy efficient but slow compared to using a pc
Anonymous No.107144020 [Report] >>107146458
>>107143875
That only applies to x265/libvpx-vp9. I'm talking about SVT-AV1 specifically because AFAIK preset 8 in that achieves similar compression efficiency as x265 fast. I got 20 FPS on my chinese android phone I got for free on a cartier switch encoding this full 1080p video.

https://download.blender.org/peach/bigbuckbunny_movies/big_buck_bunny_1080p_h264.mov

725 MB H264 to 95 MB AV1 is fucking nuts. I can't really tell the difference between them.
Anonymous No.107144106 [Report] >>107152771
This is my CPU-Z for anyone wondering. I'm getting 20 FPS encoding 1080p preset 8 AV1 video with just 2 fucking outdated Cortex A76 CPU cores running at 2.4 GHz.

This might actually be a really good/reliable way to benchmark phones desu. They can do their bullshit cheating for the first 30 seconds but the rest of the encode will force it to throttle down to true sustained performance levels.
Anonymous No.107144133 [Report]
>The ARM Cortex-A76 was released on June 1, 2018, alongside the Mali-G76 GPU. It was unveiled in 2018, with a launch date of June 1, 2018, according to sources like GSMArena.com and Reddit.
Dam, no wonder it was free.
Anonymous No.107144412 [Report] >>107148450
>tfw SVT-AV1 is faster than calculating VMAF

I got a score of 93% So a 7% reduction in quality in exchange for a 85%? reduction in filesize. Shiiieet.
Anonymous No.107145376 [Report] >>107145779
>>107143844 (OP)
No. I use the RDP app Remote Desktop Manager along with the OpenSSH Magisk module to encode videos to AV1, and then wirelessly transfer the AV1 files to my phone.
Anonymous No.107145779 [Report] >>107145802 >>107146032
>>107145376
since whend oes fucking RDP work on linux
Anonymous No.107145802 [Report] >>107145815
>>107145779
t. under a rock liver
Anonymous No.107145815 [Report]
>>107145802
enlighten me
Anonymous No.107146032 [Report] >>107146046 >>107146249
>>107143844 (OP)
what the fuck is a .mov?

>>107145779
KRdp is basically *the* standard way to do remote desktop on KDE. similar situation with Xrdp on X11 in general. the only reasonable alternative is Sunshine with Moonlight as a client, and that has way less integration.
Anonymous No.107146046 [Report]
>>107146032
x11 remote desktop is such a shitshow especially since it's supposed to do it natively (which it does but it's not well supported and kinda shit)
Anonymous No.107146249 [Report] >>107152547
>>107146032
>what the fuck is a .mov?
Nth video format nobody asked for made by Apple and they would rather crucify everyone at the stake than use sane universal things (ie .webm) like most decent human beings.
Anonymous No.107146458 [Report] >>107147693 >>107148085
>>107144020
Snapdragon 8 Gen 2
$ ffmpeg -i big_buck_bunny_1080p_h264.mov -c:v libsvtav1 -preset 8 -crf 40 -c:a libopus -b:a 128k av1.webm
...
Svt[info]: SVT [version]: SVT-AV1 Encoder Lib v3.1.2
Svt[info]: SVT [build] : Android (13624864, +pgo, +bolt, +lto, +mlgo, based on r530567e) Clang 19.0.1 64 bit
Svt[info]: LIB Build date: Aug 25 2025 00:18:31
...
[out#0/webm @ 0xb400007c7b46e300] video:84076KiB audio:8186KiB subtitle:0KiB other streams:0KiB global headers:0KiB muxing overhead: 0.295184%
frame=14315 fps= 50 q=40.0 Lsize= 92534KiB time=00:09:56.41 bitrate=1271.0kbits/s speed=2.06x
Anonymous No.107147693 [Report]
>>107146458
Hmmm that's like 2500u PC territory I think. Meh.
Anonymous No.107148085 [Report] >>107148130 >>107148316 >>107148333 >>107148450
>>107146458
>crf 40
>*double take*
>CRF FUCKING 40
disregard this fucking moron

>I don't see a difference
yeah, on your fucking phone where 1080p might as well be 480p
Anonymous No.107148130 [Report] >>107148333
>>107148085
>libsvtav1 crf default is 35
well nvm, I guess they just decided to take weird default values... again
Anonymous No.107148316 [Report]
>>107148085
I'm not op, I just ran the same command as >>107143844 (OP) to make the performance comparison fair
Anonymous No.107148333 [Report]
>>107148085
>>107148130
Comparing CRF scales across cancerous video codecs like H264 and non-cancerous ones like AV1 is not possible because mpeg-le spent every last penny they could pinch into "ORIGINAL DONUT STEEL!" patent trolling.

AV1 CRF scale goes like this:
CRF 50 = good enough quality for phone screens
CRF 40 = good enough for tablet screens
CRF 30 = good enough for 4K TVs
Anonymous No.107148450 [Report]
>>107148085
See >>107144412

VMAF scores above 90 are generally good as far as I know.
Anonymous No.107151082 [Report]
>>107143844 (OP)
Fucking love av1
Anonymous No.107151904 [Report] >>107152771
How do I try this on my supposedly RDNA3 phone?
Anonymous No.107152547 [Report]
>>107146249
mov dates back to 1991 my dude (hint: it's far older than webm or matroska which webm is derived from)
Anonymous No.107152771 [Report] >>107152958
>>107143844 (OP)
>>107144106
5700x3D 3000MT/s CL15
ffmpeg -i big_buck_bunny_1080p_h264.mov -c:v libsvtav1 -preset 4 -crf 40 -c:a libopus -b:a 128k av1.webm
frame=14315 fps=175 q=40.0 Lsize= 92535KiB time=00:09:56.41 bitrate=1271.0kbits/s speed=7.28x


>>107151904
I think you would rather pierce your own eyeballs than setup mesa-vaapi on andr*id.
Maybeee ffmpeg supports some kind of Android specific hw encoding api.
Anonymous No.107152958 [Report]
>>107152771
Yeah well, that's 175 FPS at like 100 watts. A better comparison would probably be a GPU AV1 encoder but then again those things have awful compression efficiency.