DeepAI
AIGenerate, edit, and enhance images through DeepAI's single-call REST endpoints. Agents produce and clean up visual assets inside a workflow that already knows what the asset is for.
What This Integration Enables
DeepAI is the connector you reach for when an asset arrives mid-flow and needs exactly one thing done to it. There is no job queue and no polling loop. Every endpoint is synchronous and returns a hosted output URL that the next action consumes directly, which means image work stops being a separate pipeline and becomes a step. Remove Background cuts a subject out onto transparency for product shots and portrait compositing. Colorize Image infers plausible tones for a black and white photograph. Upscale Image runs super resolution to rescue something shot too small or too soft. Generate Image and Edit Image cover the cases where the asset needs to be created or altered rather than repaired, with a negative prompt available to steer unwanted features out of the result.
That synchronous shape is a real design choice with a real cost, and it is worth naming. A flow built here is simple, but it holds a connection while the work happens, so it suits single assets and modest batches rather than a thousand-frame archive run. API access requires a DeepAI Pro subscription. Generate Image dimensions must fall between 128 and 1536 pixels in multiples of 32. These are the kinds of constraints that matter when an agent is deciding whether to call this connector or route the work elsewhere, and a connector that states them plainly is easier to build a reliable flow on than one that hides them.
Without FlowRunner
With FlowRunner
Use Case Scenarios
Supplier photography cleaned on arrival
Supplier photos arrive by email as attachments, shot against whatever background the warehouse had. An intake agent pulls each attachment, calls Remove Background to put the product on transparency, and writes the returned URL to the draft listing in WooCommerce. Before it publishes anything it checks each source against the storefront's minimum resolution. Most pass. The handful that do not go to a merchandiser with the resolution gap stated, because the decision about how to close that gap is not the agent's to make.
Archive photographs prepared for an anniversary campaign
A marketing team is building a company history page from scanned black and white photographs. The agent runs Colorize Image on each scan to produce a color version, and Upscale Image on the ones scanned at too low a resolution for the layout. Both outputs go into the page draft in WordPress alongside the untouched original, and the caption template carries a line noting the image was colorized. Presenting the original next to the processed version is not decoration. It is what lets the editor and the reader tell the difference between a photograph and an interpretation of one.
Targeted replacement inside an approved layout
A seasonal banner is approved except for one element that has to change: a product on a table needs to become the new SKU. Rather than regenerate the whole composition and lose everything else that was signed off, the agent supplies the approved banner, a mask marking the region, and a description of what belongs there, and calls Replace Image Content. The rest of the frame is untouched. The revised banner goes back to the same approver who signed off the original, with the mask attached so they can see exactly what changed.
Human-in-Loop Highlight
There is a line here between recovering detail and inventing it, and DeepAI names both sides of it in its own operation set. Upscale Image runs super resolution to clarify and sharpen what the sensor actually captured. Creative Upscale Image uses a generative model that invents plausible new detail rather than only sharpening what is already there, and the vendor's own description says it deviates further from the original. On a marketing background that is a stylistic preference. On a product photograph it is a claim about a product a customer is about to buy, made by a model that has never seen the product. So the agent runs the truthful transforms unattended and stops at the inventive ones. When a shot falls below the storefront minimum, the message reads: "SKU 4471 is 640px wide against a 1200px minimum. Standard Upscale Image will sharpen it and it will still look soft. Creative Upscale Image will fill in weave and stitching detail that is not in the original file. Which do you want on a listing image?" Replace Image Content and Colorize Image sit behind the same gate for the same reason. A person decides how much of a published image is allowed to be a guess.
Agent Capabilities
8 actionsGeneration and Editing
3- Generate Image Creates an image from a text description. Dimensions must be between 128 and 1536 pixels in multiples of 32, with recommended standard sizes such as 1024x576 and 512x512 and HD sizes such as 1216x832 and 1024x1024. The generator version trades speed for quality, and a negative prompt steers unwanted features out of the result. Returns a hosted URL.
- Edit Image Edits an existing image according to a text instruction, for example changing a subject's clothing or altering the scene. The Genius and Super Genius generator versions produce higher quality, more detailed edits. Returns a hosted URL.
- Replace Image Content Replaces the masked region of an image with new content described in text. Supply a mask image marking the area to replace, the source image, and a description of what should appear there. Used for targeted changes inside a layout that has already been approved.
Cleanup
2- Remove Background Cuts the main subject out of a photo and removes everything behind it, returning the subject on a transparent background. Used for product shots and portrait compositing where the asset has to sit on an arbitrary backdrop later.
- Colorize Image Adds color to a black and white photograph, inferring plausible tones for skin, clothing, foliage and sky. The tones are inferred rather than recovered, which is worth surfacing wherever the output is published next to historical material.
Upscaling
3- Upscale Image Runs super resolution to clarify, sharpen and enlarge an image without losing its content or defining characteristics. Well suited to rescuing a photograph that was shot or saved too small.
- Creative Upscale Image Upscales using a generative model that invents plausible new detail rather than only sharpening what is already there. Produces more striking results than standard super resolution but deviates further from the original, which is exactly why it belongs behind an approval step on anything a customer will judge a purchase by.
- Upscale Anime Image Doubles the size of an image while reducing noise using a model trained largely on anime-style artwork. Performs well on illustrations and cartoons and remains usable on photographs and other imagery.
Frequently Asked Questions
What can FlowRunner do with DeepAI?
FlowRunner agents can run Generate Image, Edit Image, and Replace Image Content in DeepAI, plus 5 more actions.
Does connecting DeepAI to FlowRunner require OAuth?
No. DeepAI connects to FlowRunner with an API key, no OAuth flow required.
Can DeepAI trigger a FlowRunner workflow automatically?
DeepAI doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.
Start building with DeepAI
$100 in credits. No card required. Connect in minutes.