Looking for creative AI-based workflows to automate website banner image editing — specifically handling text replacement and character image modifications without degrading image quality

[i_isjung]

What is your goal?

We manage a company website and regularly upload banner images that contain a mix of character illustrations and text overlays. Currently, every edit request goes through manual Photoshop work, which is time-consuming and inefficient.
The ultimate goal is to build an AI-powered image editing/automation pipeline that can handle:

Text modification requests — e.g., changing copy inside a banner without distorting the surrounding visual elements
Partial image edits — e.g., swapping out a character, adjusting a specific region, or replacing a background element
Batch or semi-automated production — reducing manual designer intervention for recurring, templated banner updates

Ideally, this would plug into an automation tool like Make.com or work as a standalone AI-assisted workflow that non-designers can operate with minimal friction.

What is the problem & what have you tried?

Core problem: Banner images contain layered compositions — character illustrations + text — but we only have flattened image files (no source PSD or editable layers). When an edit request comes in, there’s no clean way to isolate and modify just the text or a specific element.
What we’ve tried:

Canva — When replacing text inside an uploaded image, Canva either distorts the surrounding pixels or the font/style doesn’t match the original, making the result look inconsistent or unprofessional.

What we’re looking for (open to creative suggestions):

Is there an AI tool or workflow that can detect and replace text within a flattened image while preserving the background and character layers cleanly? (e.g., inpainting-based text swap)
Can tools like Adobe Firefly, Ideogram, or DALL·E inpainting handle this reliably for production-level banner work?
Would a layer reconstruction approach — recreating banners as templated assets from scratch using a design system (Canva API, Bannerbear, Figma tokens) — be more sustainable than trying to edit flattened images?
Is there a way to set up a Make.com automation where an edit request (e.g., via form or email) triggers AI image editing and outputs a revised banner automatically?
Any recommended approaches for maintaining brand/style consistency when AI regenerates parts of a banner image?

Error messages or input/output bundles

No runtime errors — this is a workflow design and tooling question. The problem is upstream: we lack an editable source file (layered PSD/Figma), so all AI editing attempts work on flattened raster images, which leads to quality degradation when text or elements are modified. Looking for the most practical path forward given this constraint.

[mszymkowiak]

Hello,

@i_isjung

Text modification requests — e.g., changing copy inside a banner without distorting the surrounding visual elements

Do not use AI for this.
It can be achieved using, for example, PlacidApp or GenerateBanners, which are well integrated with Make and provide 100% accuracy as they work on input and pre-defined text layouts.

@i_isjung

Is there an AI tool or workflow that can detect and replace text within a flattened image while preserving the background and character layers cleanly? (e.g., inpainting-based text swap)

According to my knowledge and experience- no. There is currently no tool that can do this in a production-ready way- meaning it will generate exactly what you want every time without issues.

A better approach is to generate (or create) background images and process them separately as banners with text.

@i_isjung

Can tools like Adobe Firefly, Ideogram, or DALL·E inpainting handle this reliably for production-level banner work?

No.

@i_isjung

Would a layer reconstruction approach — recreating banners as templated assets from scratch using a design system (Canva API, Bannerbear, Figma tokens) — be more sustainable than trying to edit flattened images?

Yes. As described earlier, you can try using AI to modify the image and then process it with tools designed for adding text.

@i_isjung

Is there a way to set up a Make.com automation where an edit request (e.g., via form or email) triggers AI image editing and outputs a revised banner automatically?

Yes.
You can use Mailhooks which work as dedicated inbox to which you send emails, or- to handle input more reliably (dividing image prompt instructions and banner text/ text generation prompt instuctions), connect a form service to Make. For example, Tally.so, which offers a generous free plan.

@i_isjung

Any recommended approaches for maintaining brand/style consistency when AI regenerates parts of a banner image?

AI is, by nature, non-deterministic. If you want strict brand consistency, do not rely on AI-generated results alone.
AI can produce good results, but in a production environment it should be used in a controlled workflow:

1. Use clean template images.
2. Apply AI only to those templates.
3. Add brand elements, filters, or other visual components using dedicated tools.
4. Once the results are approved, add the text layers.

[i_isjung]

Oh, thank you for adivice. I’m gonna try.

Your idea was a great help to me.

Thank you.

[DavidGurr_Make]

@mszymkowiak

A better approach is to generate (or create) background images and process them separately as banners with text.

This, x100!

Placid is a great solution for this, however as you mentioned Canva if you have a Canva Enterprise plan you can use the “Autofill a Design from a Brand Template” module.