>>510485039That is exactly where we're going
the thing is, software giants are desperate to build these services behind cloud paywalls
but the open-source community, specially china, is giving away all these tools for free, albeit with terrible horrible interfaces, but it's photoshop vs GIMP or 3DMax-Maya vs Blender all over again
for now, powerful models (like the ones used for Veo3) can only be run by very powerful GPU clusters, but it doesn't require insane amounts of GPU/VRAM either, although it is close to triple digits (because we can see the requirements in the papers/github), which is obviously too expensive for a regular user
now, if the new AI chipsets boosts the capacity and for example x2, x4, x8 the computational power AI has using gaming GPUs, then we're off to the races, because the equivalent of a powerful gaming GPU could represent close to triple-digits VRAM or even more, and then you would be able to run something like Veo3 on a gaming rig