Ehhh. Mixed bag. Integrating AI into every-damn-thing is half of why people act like a robot kicked their dog - but diffusion is the half that unambiguously does what it's supposed to, and a popular FOSS tool is a decent place to offer an it-just-works installation for local models.
The images are generated on volunteer GPUs through AI Horde.
Nevermind, fuck this.
Thanks! Hm, honestly not seeing anything amazing that stands out tbh.
Here's a comparison from r*ddit. Hopefully people are able to fine-tune on this. It's got a way better license than Flux.
Interested to see how it compares to flux, if it's not done already. Most of my workflows are fairly flux native, but I'll convert if it's worth it
Okay, you let me tie this into a soreadsheet ir something to geberate charts, and there's finally a use case for this that i like.
Im not sure it's worth needing a 5080 to make ultra pretty graphs, but, you know; smoke em if you got em.
This is the first time I’ve seen a model target SVG drafting. Anything you have seen previously about unicorns or whatever was just someone experimenting with interesting edge case usage of models not designed for this purpose.
Feeding a language model a bunch of vector art does not seem productive to me. So it makes sense that something like GPT4 sucks at it.
Just gave it a try. I couldn't get coherent results from img-to-svg with a few different tests of low-res pixel art and high-res cartoons. txt-to-svg also gave me incoherent blobs even with simple prompts. Something must be wrong there. Is it working for anyone else?
I might just try installing it locally when I get home.
The project page didn't have a link to it, but there is a demo on HF.
Hard to judge quality when what we're seeing is practically a pixel-perfect recreation. The tricky part of automated vectorization is detecting and plotting curves in such a way that it scales correctly. Bad implementations will use too many elements, or include straight lines that should be parts of curves, etc. Those errors would not be visible in those low-res rasterizations.
I want to fine tune this model on large geospatial datasets.
Your name is crazy. 🤣
It can do IMG to SVG. Check out the right side of this image:
[creams self in simple features]
I am very into this if it can take a non-vector graphic as input and work to that. OpenAI's attempts at that have been complete dickfarts
Comfy UI isn't nearly as easy to use as think it is. It's literally the hardest to use of the GenAi UIs. It's not lowering the cognitive barrier to anything.
Lowering the cognitive barrier for use is not a feature in this political climate — especially when anti-intellectualism is a Fascism 101 pre-req. 🤢
This isn't anything you can't already do easier on Photoshop. This is just a comfy UI implementation.
Don't worry, folks! That's not going to immediately be used for political fraud & culture-wide psyops. They promise.
- Reporting. As part of Your use of a Licensed FLUX [dev] Model, you must provide us certain data related to your use of the Licensed FLUX [dev] Model or Derivatives, such as the number of images that are generated by the Licensed FLUX [dev] Models or Derivatives, or we may require you to use software to track and report usage to us. You agree to provide us those metrics or use any required software to collect that data for us.
So .. I assume comfyui ain't going to do this
That is amazing. I have got to get my GPU water cooler done so I can do video. How this is interacting with the prompt in the model loader code seems intriguing to dig into
Oooh, sparrrkly. 🤩🤘🏼
Bets on how long before BambuLabs acquires it?