ComfyUI 10 Comparison of 19 Checkpoints

Link to my workflows used in YouTube videos:

19 Checkpoints render an image with the same prompt and the same parameters. The results are a similar image yet with subtle differences. Ay yje end of the video the 6 checkpoints that render the most realistic photo are shown in one view. To compare the checkpoints yourself, use a workflow like for instance ConfyUI default, or use 200 Comfy Lora2 from the json folder in the 'A lot of unsorted workflows' link above.

Pos Prompt: front shot, portrait photo of a happy american woman, age 20, natural skin, her brown hair in waves, skin moles, thin golden necklace, old cream cracked cement wall background
Neg Prompt: (worst quality, low quality, illustration, 3d, 2d, painting, cartoons, sketch), open mouth
Size: 1024 x 1024
Seed: 931425002069238
Steps: 64
CFG: 4.5
Sampler: dpmpp_3m_sde karras

[tsmakrakis]

One of the best channels. Thank you for sharing the workflow! Great attention to detail. Your videos are really helpful!

[erperejildo]

I have a iGPU AMD 680M. In BIOS I assigned to it 16gb and I can pretty much render what I want, the problem is that some checkpoints (normally the ones around 6gb) fail. For now I'm just doing some tests, but I'd like to generate kids illustration, so nothing super fancy like super detailed faces.

Any checkpoint recommendation for that?

[DerekShenk]

Would be handy to be able to compare checkpoints however your google drive gives an error stating you deleted this workflow. Do you have a new one you can share?

[joeduffy]

That's a lot of Steps, dude! I run at about 24, depends on the sampler I suppose, though. Does it really make that much difference after a certain point?