search

LostRuins / koboldcpp

A simple one-file way to run various GGML and GGUF models with a KoboldAI UI
https://github.com/lostruins/koboldcpp
GNU Affero General Public License v3.0
4.4k stars 318 forks source link

Add images to any message (Chat)? #732

Open aolko opened 4 months ago

aolko commented 4 months ago

Can i add multiple images and can i add images to any message (including my own, not just replies)?

LostRuins commented 4 months ago

Yes you can. Click the "Add Img" button and you can add an image at any time, and you can reposition images in edit mode.

aolko commented 4 months ago

Can i trigger that via text?

LostRuins commented 4 months ago

You can set images to "Autogenerate" and it will generate a new image every few hundred words.
What do you mean by "trigger via text"?

aolko commented 4 months ago

You can set images to "Autogenerate" and it will generate a new image every few hundred words. What do you mean by "trigger via text"?

just like you'd do in gpt-4 - "draw me a...", "generate a...", "give me a photo of a..." i think united or some other gui like sillytavern or similar has that, no?

aolko commented 4 months ago

Oh and can it expand WI entries when sending the prompt to a1111?

ptr2019 commented 4 months ago

You can set images to "Autogenerate" and it will generate a new image every few hundred words. What do you mean by "trigger via text"?

just like you'd do in gpt-4 - "draw me a...", "generate a...", "give me a photo of a..." i think united or some other gui like sillytavern or similar has that, no?

I have tried "draw me a..." prompt (or ask for a SD prompt) and got a prompt for SD in return. Then the buttons "Add img - Generate Image (Automatic)" gives the desired image (wondering if there is a way more directly, though).

BTW, any SD model recommended (like those with large vocabulary or multilingual feature)? It seems that the text model generates results beyond the comprehension of the visual model...