Expose additional API endpoints to integration AI-based image generation
K
Kyle Bahr
This is a big ask, but here it goes.
I've been debating creating a RenderForm template with various elements:
- background image
- overlay graphics
- text layers
- logo placeholders
- etc
And then building an automated workflow that would essentially,
- take the initial template variables to create an image
- pass that image to AI (OpenAI, Claude, etc)
- pass a breakdown of API options (layers, positioning, sizing, etc)
- pass a prompt based on the desired outcome (e.g. take the base image and implement the logo, text layer 1, text layer 2, etc - in a way that it creates a properly designed image.)
- allow the AI to create a finalized image
Use case example:
- pass a generic "holiday" graphic and request that client logo be added where it fits best
I think this could probably be completed without any support from the RenderForm team - however it would likely take MANY API requests to get a desired result.
Open to any feedback you have on this.
Thanks
Jakub Pomykała
Kyle Bahr Regarding the original request: Sure thing, go ahead! If there is anything that would be required from the API side, then just let me know. However... I've tried to do something similar a year ago using OpenAI, and the results were not good enough, but it all depends on the prompt and what you want to get. In my case, I wanted to automate creating templates, and OpenAI didn't really get the whole idea of creating an appealing image when having details regarding variables (position, size, what it is, how to use it, etc.). OpenAI propositions looked somehow random. 😄
K
Kyle Bahr
An add-on to this.
I've been working on an ongoing issue of using AI to crop images to ensure the main subject matter is visible within a RenderForm template.
This issue is separate from the original feature request - however it is similar (i.e. allowing AI solutions to help generate better images)