I like AI process for research but I find really funny to see some artist signature sometimes... credits should go to them when you are inspired by their work ;) cheers !
Could you guys do a video on various ways of upscaling? Using the hi-res fix always radically changes the image generated without it and is so frustrating, that I end up using image to image. But there must be better ways in invoke.
Any insights on why using the IP adapter is dramatically slower than not using it for me? And I don't mean like a few percentages slower speed I mean it goes from like 2 it/s to 10 s/it from having it enabled at all. This isn't always the case either, seems like certain Invoke versions can handle the IP adapter and others can't at all on my PC.
You're probably using enough VRAM that the CUDA drives are swapping stuff to normal RAM depending on what else is using VRAM at that time, or models that InvokeAI is caching in VRAM. All that assumes you're on a NVIDIA Card and a driver version that does the swapping.
Around 55 minute mark you can clearly see some sort of artist signature for that character. Is that the SDXL base model trainer or the Juggernaught XL model your using. This nontransparent training models are okay to commercialize? I'm really curious on how your expecting professional studios to sign up for a service you guys are providing that is trained on copyrighted images where the creators really most likely did not consent to? Companies don't want lawsuits.
Invoke is becoming part of my workflow, (previously Photoshop). Keep up the great work 👏
many many thanks, love you, love Invoke
I like AI process for research but I find really funny to see some artist signature sometimes... credits should go to them when you are inspired by their work ;) cheers !
Could you guys do a video on various ways of upscaling? Using the hi-res fix always radically changes the image generated without it and is so frustrating, that I end up using image to image. But there must be better ways in invoke.
Any insights on why using the IP adapter is dramatically slower than not using it for me? And I don't mean like a few percentages slower speed I mean it goes from like 2 it/s to 10 s/it from having it enabled at all. This isn't always the case either, seems like certain Invoke versions can handle the IP adapter and others can't at all on my PC.
You're probably using enough VRAM that the CUDA drives are swapping stuff to normal RAM depending on what else is using VRAM at that time, or models that InvokeAI is caching in VRAM. All that assumes you're on a NVIDIA Card and a driver version that does the swapping.
ipadapter and t2i are less, how to install more
👋
Around 55 minute mark you can clearly see some sort of artist signature for that character. Is that the SDXL base model trainer or the Juggernaught XL model your using. This nontransparent training models are okay to commercialize?
I'm really curious on how your expecting professional studios to sign up for a service you guys are providing that is trained on copyrighted images where the creators really most likely did not consent to? Companies don't want lawsuits.
You would rather prefer let multibillion corporations to monopolize AI, so they can charge us whatever they want?