Messing around in Gemini/VEO 3

[TigerRaptor]

Been messing around with it here and there.

[vision]

Neat, I bet we saw the same article about turning a picture in to a model. Though seeing that D2 boss is making me think I should 3D print one. I was thinking about making a custom case for one of several Raspberry PIs in my house, maybe I'll make a PI that runs Descent and put it in a boss-shaped housing?

[TigerRaptor]

Yeah, there's some Descent models I would like to bring to life. 3D printing the classic fusion cannon is one of them.


Still pushing the limits in VEO, but for some reason I can’t get the eye to move. I might give Wan 2.2 in ComfyUI a try next. Hey Isaac, any ideas?

[TigerRaptor]

I'm done experimenting with this, so one more. One of Kuman's pictures brought to life.

[Isaac]

I might give Wan 2.2 in ComfyUI a try next. Hey Isaac, any ideas?

I was pretty limited by my VRAM, About 54GB of unified memory to the GPU and the rest to CPU's RAM and SWAP. It's cool what's coming out but I think the NVIDIA Spark (or digits?) might be the logical step for me to take.

[TigerRaptor]

Well, ComfyUI will be on hold for a while since they don't support AMD cards. The work around trying to get it to work is a real ★■◆●ing ★■◆● both in Windows and Linux.

[Isaac]

It's probably for the best as there's no FOSS model on the same level for making video as Sora 2 or VEO 2. But that might change with new versions of wan coming out.

[TigerRaptor]

One thing that annoys me about VEO sometimes is that it doesn’t always follow directions. For example, when I asked it to fire five fireballs, it unleashed an entire barrage instead.

[Isaac]

That's still pretty cool!