- This topic has 4 replies, 2 voices, and was last updated 1 year, 3 months ago by
.
-
Posts
-
I just ran across this today on Reddit, downloaded a gguf model and default workflow, haven't tried it out yet. Pick the model that fits your VRAM: https://huggingface.co/QuantStack/FLUX.1-Kontext-dev-GGUF/tree/main
EDIT 06/29/2025: fixed the above link, had FLUX-only checkpoints before.
It takes multiple 3d (i.e. Daz) poses and (real) target images, then spits out a single image incorporating all the poses. Text-prompt any details to include/exclude.
Basically it saves you from separate inpaint/outpaint, regional-prompts, control-net, ip-adaptor configs and just 'does it all' to give you a realistic finished scene based on the 3d scene-pose you feed comfy, or so it says. Reddit says you might want to add a face detailer and upscale node to the default workflow.
Anyway, looks interesting so thought I'd pass it on.
https://civitai.com/models/1722303/kontext-character-creator
I read it through and it looks darned interesting. People seem to be having good results adding or removing people.
On a similar note, I don't 'get' HuggingFace. How do you know what the items are when there's no pictures?
Thanks!Yeah, I know. Huggingface is just text, so you have to already know what you're going for. I got that link off Reddit or I would have ended up with a model that wouldn't fit my meager 3060-12GB instead of having a choice of quantized models.
The new Comfy 3.43 just added internal support for Kontext and gguf like three days ago, those guys are right on top of it. Stuff changes so fast in SD, but I guess this is the next new thing, at least for a week or so...
*chuckle* I wouldn't call a 12GB 3060 meager. I went from an ancient 920 card (or something) with 12 Kb's of ram (or so it felt like) to a 3060 when NFTs and crypto collapsed. It still feels like a rocket to me!
EDIT 06/29/2025: corrected my original link in 1st post
https://huggingface.co/QuantStack/FLUX.1-Kontext-dev-GGUF/tree/main
These are the Quantized, low-RAM, gguf checkpoints for Kontext.Edit: Just saw your reply. I have no problem with 3060, it's great for rendering. But the way SD is going, the current accepted low-mem card is 16GB and even 24GB cards are having trouble with some checkpoints. The good thing is, sometimes there are low-RAM Quantized models available.
- You must be logged in to reply to this topic.
