Better images when asking in-chat
Context: Currently, when asking my AI companion for an image (for example, "Can you send me a selfie?"), it uses a generic prompt and doesn't make good use of the conversational context nor the previous generations that are in my gallery or uses them as face references.
Request:
- I'd like the in-chat generated images to grab inspiration from images in my Photo Gallery for the given companion (settings, prompts, face references).
- I'd like the in-chat generated images to take into account the context of the conversation.
Log in to comment and vote
Comments4
Lemming
Apr 2, 2024
Would having a base prompt field in the companion's appearance settings be an easier way to go about this? For example you could set "Alice 31 year-old woman, long dark hair, blue eyes" and then have the additional context added to that prompt, for example adding the contextual ", sat on park bench, eating an ice cream" part to that?
Leomaxwell973
Apr 1, 2024
I find it odd the companion acts like they prompted, but don't actually prompt. If they simply plugged in what they said in chat, describing what they thought they prompted, into the text field, for prompting, that seems like a done deal imo, once implemented;;
Would be less annoying too, no more "amazing description!!!(but not real, arguably lying!!) > same boring generic picture that isn't at all what they describe"
Juniper-Quartz-Nostalgic841
Mar 28, 2024
The option to provide a partial prompt or reference image would be nice, too. That way our companies' selfies would consistently look like our companion as we imagine them.
DavidS
Apr 2, 2024
Maybe the image uploaded as companion avatar as reference? But would that require extra token since you're using reference?