Subnostr

Replying to

JSTR

Another options is dall-e for a paid version, or stable journey for self host… but you have to have a good graphic card

HoloKat 2y ago

Stable diffusion is not bad and can make some great images too.

Reply to this note

Please Login to reply.

Discussion

JSTR 2y ago

That’s what I’m using… midjournry has a very distinct look… shiny I would say… I have to fiddle a lot to get something I’m vibing on… is midjourney easy to get a good result?

HoloKat 2y ago

If you want truly realistic humans; you have to train a custom model with stable diffusion. Lexica.art has their own model for realistic humans that looks really good.

JSTR 2y ago

Oh this is awesome, Ty!

Dr.EAD 2y ago

I think I’ll just go hug a real person first, but if I find the time or that doesn’t work out this’ll be the next thing I try. 😌😉😂

?cid=2154d3d7cuj1k15ythr0zxtsdba0ehhzkq07jk2yxdisbrdd&rid=giphy.gif&ct=g

Jorj X McKie 2y ago

Haven't tried that one... can it insert text into an image? MJ completely fails at text.

JSTR 2y ago

Stable diffusion is really poor at text… at least 1.0 is

Jorj X McKie 2y ago

I can understand why AI 'art' struggles with hands/fingers, but training it for simple language characters-fonts & and restricting blacklisted words-phrases seems low bandwidth to me, but I'm told it's a 'high-CPU' thing. 🧐