EliteMooseTaskForce
tomcat 
I'll be watching this and hoping for a merge!
VAAPI Encoder (2nd branch/attempt) by sgothel · Pull Request #7467 · HandBrake/HandBrake · GitHub
#handbrake encoding with default #Mesa stack does not utilize GPU acceleration without proprietary amf-amdgpu-pro drivers, which are available in #AUR but look to have been abandoned after AMD stopped publishing separated AMF from their stack, building on top of open amdgpu drivers instead, and offer inferior performance in other contexts.
#ffmpeg can with #VAAPI but hammering out scripts for the amount of media I'm archiving, that's less than ideal.
Could spin up a VM with amf drivers, but setting up with SingleGPUPassthrough seems like overkill/pain in the ass.
Any other tools utilizing VAAPI that could make things a little easier?
hoergenOBS, Flatpak & VAAPI Solution
#Obs #Streaming #Linux #Opensource #VAAPI
hyperblog.de/hoergen/2026/04/1…
Es gibt bei Linux OBS der Flatpak Version und der installierbaren Version wohl ein Problem mit dem Hardware Rendering bei einigen Computern.
OBS, Flatpak & VAAPI Solution - Dem hoergen Blog
Es gibt bei Linux OBS der Flatpak Version und der installierbaren Version wohl ein Problem mit dem Hardware Rendering bei einigen Computern. Das Problem hatte ich auch mit einem etwas älteren Computer mit einer integrierten Intel Grafikarte. Obwohl die nonfree Inteltreiber installiert sind (Info weiter unten). Auf einem anderen Computer habe ich dieses Problem nicht. … „OBS, Flatpak & VAAPI Solution“ weiterlesen
boredsquirrelComment on this video (which us not shown on my Fedi instance, yay federation!)
https://tube.kockatoo.org/w/63cGWTvhzqu2bwMGsh3QcF
I cannot believe that that GPU is not usable for video work. That is what GPUs are made for!
Poorly, there are a few caveats:
- there are always more DEcoders supported than ENcoders. I guess not being able to play a video (it would play but be less efficiently decoded on the CPU) is worse than being bound to #h265 instead of #av1 for recording.
- thus it makes sense that GPUs generally work flawlessly for gaming, as that is just displaying videos (or rendering content, which is kinda different, no idea why video codecs are more difficult than rendering methods)
- for every video format out there exists an encoder. If you use software rendering (on the CPU), you can use ANYTHING. Like encode to modern (and free!) AV1 using #svtav1 on any old CPU.
- if you want to encode videos on a GPU, you need drivers, but also a different encoder. On #NVIDIA you use #NVENC (NVidia ENCoder), for example nvenc_x265 for h265/HEVC videos. On #IntelArc (and afaik AMD GPUs), you use the #QSV (quick sync video) library which builds on top of #VAAPI, which is open source. So you can encode to AV1 using qsv_av1 for example.
What encoders are supported depends on the GPU a lot. Intel Arc seems to have supported the free, efficient and future-proof AV1 codec the earliest, but also the older free codecs VP9 and VP8 (also used in webm), like in a GPU I recently bought used. My older NVIDIA GPU only supports H265, which is proprietary and often less efficient than AV1 (not always, video codecs are black magic).
As OBS can use your GPU for video capture (encoding), the issue is purely in software support.
I only have experience with #ffmpeg for re-encoding videos, but that is also in many video editing tools. Try if you can get them somehow.
For NVIDIA I needed the latest proprietary drivers (#NixOS makes this extremely easy, can recommend) and ffmpeg, nvenc worked out of the box.
For intel Arc I used the regular drivers included in the kernel, but after adding the QSV runtime to my "hardware.graphics.extraPackages" I needed to recompile the entire kernel. Using a #longterm kernel (currently 6.18) makes this more viable... longterm XD.
My Childhood Dream Laptop
Ethan BlantonWhen I try to use hardware video encoding on my Intel ARC 380 with ffmpeg and VAAPI, it keeps hard locking the whole system. It seems like it might be related to the mouse cursor, but maybe that's a spurious association. I'm running Debian trixie with the backport kernel (6.18.12+deb13) and 20250410 Intel firmware. If anyone has ideas for making it stable, I'm listening. (I've also had occasional hangs with firefox and hardware media decode, so I disabled that!) #askfedi #vaapi #ffmpeg
PsychoticSheepI tried Qwen2.5-Coder-7B-Instruct.Q6_K locally with Ollama as the loader, asking it to create a simple Snake game in Python with Pygame and as an extra challenge, the instructions were given in German.
The game works well: the snake grows correctly, the grid and colors are fine. I just had to give the model a little nudge in two places:
- Don’t change the food color every frame
- Avoid recursive gameLoop() for "Play Again"
Qwen2.5 is a great co-pilot that handles most of the work, leaving only minor bugs to correct. German works surprisingly well ("Schlankkörpers" instead of "Schlangenkörper" does not matter, such errors can also occur in large models from time to time...) even though the main language is English. The model supports many programming languages such as: Python, C, C++, Java, JavaScript, HTML/CSS, Bash, SQL… and more.
Conclusion: It still doesn't work completely without programming knowledge, but as a local assistant Qwen2.5-Coder is excellent.
btw my prompt was: "write the game again."
Video workflow:
- Recorded with OBS
- Edited in Kdenlive
- Transcoded with VAAPI (H.264)
No cloud, real hardware.
Everything runs on Linux + Text Generation Web UI (FOSS), so anyone can set this up.
No GPU? No problem, you can also run it using PyTorch’s CPU backend, just much slower.
Background music: ALICE - CROSS THE BORDER (https://www.youtube.com/watch?v=dcqbWgxW4oU)
#Qwen2 #LLM #LocalAI #Ai #vibecoding #Python #Pygame #CodingAI #FOSS #Linux #SnakeGame #Ollama #AIcoPilot #MultilingualAI #TextGenerationWebUI #OBS #Kdenlive #VAAPI #NoCloud #LocalAIWorkflow
The game works well: the snake grows correctly, the grid and colors are fine. I just had to give the model a little nudge in two places:
- Don’t change the food color every frame
- Avoid recursive gameLoop() for "Play Again"
Qwen2.5 is a great co-pilot that handles most of the work, leaving only minor bugs to correct. German works surprisingly well ("Schlankkörpers" instead of "Schlangenkörper" does not matter, such errors can also occur in large models from time to time...) even though the main language is English. The model supports many programming languages such as: Python, C, C++, Java, JavaScript, HTML/CSS, Bash, SQL… and more.
Conclusion: It still doesn't work completely without programming knowledge, but as a local assistant Qwen2.5-Coder is excellent.
btw my prompt was: "write the game again."
Video workflow:
- Recorded with OBS
- Edited in Kdenlive
- Transcoded with VAAPI (H.264)
No cloud, real hardware.
Everything runs on Linux + Text Generation Web UI (FOSS), so anyone can set this up.
No GPU? No problem, you can also run it using PyTorch’s CPU backend, just much slower.
Background music: ALICE - CROSS THE BORDER (https://www.youtube.com/watch?v=dcqbWgxW4oU)
#Qwen2 #LLM #LocalAI #Ai #vibecoding #Python #Pygame #CodingAI #FOSS #Linux #SnakeGame #Ollama #AIcoPilot #MultilingualAI #TextGenerationWebUI #OBS #Kdenlive #VAAPI #NoCloud #LocalAIWorkflow
Small making-of of my “True Beauty Is So Painful” piece (listening to “True Beauty Is So Painful” by Oomph! in the background), because “AI art = just pressing a button” is still a thing.
Here I’m showing briefly (15 MB max file upload) my SDXL workflow in ComfyUI, from node structure to model choice to parameters.
LoRAs in this setup are only linked to the positive prompt, because I wanted to fine-tune their weights there specifically, without affecting the negative prompt.
During rendering, I ran in parallel:
- GPU load with radeontop, you can clearly see how on RDNA2 everything (matrix multiplications, convs, etc.) runs over the shaders
- Temperatures & power states briefly shown with corectrl
Peak at 187 W, hotspot briefly at 97 °C
RDNA2 doing RDNA2 things…
Video workflow:
- recorded with OBS
- edited in Kdenlive
- transcoded with VAAPI (H.264)
No cloud, just decisions, iteration and real hardware.
Everything runs on Linux + ComfyUI (FOSS), so anyone can set this up.
No GPU? No problem, you can also run it using PyTorch’s CPU backend, just much slower.
#AIArt #ComfyUI #SDXL #stablediffusion #LoRA #FOSS #Linux #AMD #RDNA2 #GPUComputing #OpenSource #AIWorkflow #OBS #Kdenlive #VAAPI #DigitalArt #MakingOf #AIProcess #NoCloud
Here I’m showing briefly (15 MB max file upload) my SDXL workflow in ComfyUI, from node structure to model choice to parameters.
LoRAs in this setup are only linked to the positive prompt, because I wanted to fine-tune their weights there specifically, without affecting the negative prompt.
During rendering, I ran in parallel:
- GPU load with radeontop, you can clearly see how on RDNA2 everything (matrix multiplications, convs, etc.) runs over the shaders
- Temperatures & power states briefly shown with corectrl
Peak at 187 W, hotspot briefly at 97 °C
RDNA2 doing RDNA2 things…
Video workflow:
- recorded with OBS
- edited in Kdenlive
- transcoded with VAAPI (H.264)
No cloud, just decisions, iteration and real hardware.
Everything runs on Linux + ComfyUI (FOSS), so anyone can set this up.
No GPU? No problem, you can also run it using PyTorch’s CPU backend, just much slower.
#AIArt #ComfyUI #SDXL #stablediffusion #LoRA #FOSS #Linux #AMD #RDNA2 #GPUComputing #OpenSource #AIWorkflow #OBS #Kdenlive #VAAPI #DigitalArt #MakingOf #AIProcess #NoCloud
NevyI just spent like 5 hours trying to force #FFMPEG #VAAPI #OBS encoder to output something that isn't a hot blurry mess. It turns out there is a nice ffmpeg webpage, where they state that AMD/Mesa platform supports one parameter:
And setting it to 29 activates a quality preset which isn't complete bullshit and then the #stream looks nice...
https://trac.ffmpeg.org/wiki/Hardware/VAAPI
I love #Linux, but cut me some slack sometimes xD I will start making guides for all of the bullshit I have been finding out about stuff over the years. The solutions people post on forums are sometimes so ridiculous xD
At 1:40 AM, Nevy over and out... Finally... 
-compression_level=xxAnd setting it to 29 activates a quality preset which isn't complete bullshit and then the #stream looks nice...
https://trac.ffmpeg.org/wiki/Hardware/VAAPI
I love #Linux, but cut me some slack sometimes xD I will start making guides for all of the bullshit I have been finding out about stuff over the years. The solutions people post on forums are sometimes so ridiculous xD
At 1:40 AM, Nevy over and out... Finally...
BlablaLinux