https://www.youtube.com/watch?v=RAzbzMgVA2A
So I’m not really across image generation very much, but this popped up on my YouTube feed today and I was blown away.
I was wondering if anyone had any idea of the kind of hardware and LLM they may have used to create this. Surely it must be cloud based and not open-weight, local llm?
I’m aware that a few months ago, a new, very capable Qwen image/video/audio generator was released, but I can’t imagine what was used for this. Is this the new bytedance or something else? I would be curious in dabbling with it to make something fun for my kiddos.


That’s gonna be something like Veo 3.1 right?
I genuinely don’t know! I know there was an influx of slop with the most recent Seedance some months ago, but this seems to be a cut above.
It’s the first I’ve watched one of these and had to do a double take. Clearly the person that put this together spent quite a lot of effort in staging it, figuring out the transitions, prompting the LLM. It’s not just “Claude; make the new Avengers movie - no mistakes”.
Does anyone here generate videos like this? And if so, would you mind speaking a little bit about the workflow, tools etc? I imagine you’d need an ungodly rig to get something like this made. Either that or a lot of time / batch processing.
wouldn’t be local i don’t think, i’ve not heard much regarding open source video generation