Some big news in AI image generation today.
Most of the members of the image and video team from Stability AI launched a new company called Black Forest and a new model called Flux. Early results look very good. Currently requires 24 GB VRAM, but should be optimized by the community soon. ComfyUI has support already. Their site has a teaser of the upcoming text-to-video model, but it's currently being hugged to death.
https://github.com/black-forest-labs/flux
https://blackforestlabs.ai/up-next/
JoyCaption is a free, open and uncensored VLM built from the ground up. Existing tools are expensive or don't perform well for some domains. Great results for a pre-Alpha. This tool will be extermely useful to enable more community fine-tuning and training because producing captions for datasets is very time-consuming.
https://www.reddit.com/r/StableDiffusion/comments/1egwgfk/joycaption_free_open_uncensored_vlm_early/
#stablediffusion
John Dee
npub1lc…pw464
2024-08-01 19:51:00
Author Public Key
npub1lceznr3f426wc2g3crdamfy9cpels6w9g38wjtm6ufr76gz3vfjskpw464Published at
2024-08-01 19:51:00Event JSON
{
"id": "a3f53edea934a4a1a563a9983687e1008b0406d49a909bd742a85d23ef5fe9cd",
"pubkey": "fe32298e29aab4ec2911c0dbdda485c073f869c5444ee92f7ae247ed20516265",
"created_at": 1722541860,
"kind": 1,
"tags": [
[
"t",
"stablediffusion"
]
],
"content": "Some big news in AI image generation today.\n\nMost of the members of the image and video team from Stability AI launched a new company called Black Forest and a new model called Flux. Early results look very good. Currently requires 24 GB VRAM, but should be optimized by the community soon. ComfyUI has support already. Their site has a teaser of the upcoming text-to-video model, but it's currently being hugged to death.\nhttps://github.com/black-forest-labs/flux\nhttps://blackforestlabs.ai/up-next/\n\nJoyCaption is a free, open and uncensored VLM built from the ground up. Existing tools are expensive or don't perform well for some domains. Great results for a pre-Alpha. This tool will be extermely useful to enable more community fine-tuning and training because producing captions for datasets is very time-consuming.\nhttps://www.reddit.com/r/StableDiffusion/comments/1egwgfk/joycaption_free_open_uncensored_vlm_early/\n\n#stablediffusion",
"sig": "a2e10609fdbd7e1b8e155697147a66db8e5403e3c14de177ed2335396c903377129610bb1e911d3c7e2ac5a6bb825566f2b1ed96566b9847ac3cee83e9d4281f"
}