Ideogram
AIGenerate and edit images with Ideogram v3, known for accurate in-image text rendering. Agents create visuals with real typography, inpaint with masks, remix and reframe images, replace backgrounds, upscale, and caption images, with results saved to durable FlowRunner file URLs.
What This Integration Enables
Ideogram is the model you reach for when the image has words in it. Most generation models treat text as texture and produce lettering that falls apart on inspection; Ideogram v3 renders headlines, labels, and logos as accurate typography, which is what makes generated visuals usable for promo tiles, posters, and packaging mockups rather than just mood boards. The connector adds the piece automation actually needs: Ideogram's own URLs expire quickly, so every result is downloaded to FlowRunner file storage and returned as a durable link a flow can hold, review, and publish hours later. Because a model that renders text well will also render a wrong price with perfect confidence, publication stays behind a [human-in-the-loop](/concepts/human-in-the-loop) gate. - Generate visuals with accurate in-image text, style presets, color palettes, and reference images - Edit, remix, reframe, and re-background existing assets instead of regenerating from scratch - Upscale approved picks to production resolution with resemblance and detail controls - Caption any image for accessibility text or prompt reverse-engineering
Without FlowRunner
With FlowRunner
Use Case Scenarios
Weekly offer tiles without a design queue
Merchandising drops the week's offers into [Google Sheets](/integrations/google-sheets): product, price, promo line. For each row, the agent calls Generate Image with the exact offer text in the prompt and the brand's color palette preset, then Describe Image to verify the rendered price matches the sheet before a human ever looks. Candidates post to [Slack](/integrations/slack), the marketing lead approves each tile, and the agent publishes the durable file URLs to [Shopify](/integrations/shopify) collections. The store gets typography-correct offer graphics every Monday, and the designer's queue holds design work instead of production work.
One hero asset, every placement
A product launch has one approved hero visual. The agent calls Reframe Image to outpaint it into each placement's shape: wide for the [Webflow](/integrations/webflow) banner, square for the feed, tall for stories. Where a placement needs different text, Edit Image inpaints the masked headline region with the localized line while leaving the artwork untouched. Each variant is upscaled, captioned with Describe Image for alt text, and staged for review. The campaign ships one visual identity in six shapes without six separate design tickets.
Product photos that survive the marketplace
Supplier photos arrive inconsistent: mixed backgrounds, low resolution, no styling. The agent runs Replace Background to put every foreground subject on the catalog's standard backdrop, Upscale Image to lift resolution to marketplace requirements, and Describe Image to draft the accessibility caption. Results land in a review grid, a merchandiser spot-checks that no foreground got mangled, and approved shots publish to [BigCommerce](/integrations/bigcommerce) with their captions attached. Remix Image handles the seasonal variants when the same product needs a holiday treatment.
Human-in-Loop Highlight
Ideogram's signature strength is the risk. Because Generate Image and Edit Image render words as crisp, confident typography, a wrong number looks exactly as authoritative as a right one: a tile reading $1,999 when the sheet says $2,999 is not a blurry artifact anyone would catch, it is a legible price a customer can screenshot. So the publish step is never automatic. Before any durable file URL leaves the flow for a storefront or a scheduler, the agent presents the rendered image beside the source copy it was generated from, and a person confirms every word and digit matches. The generation is cheap to redo. The screenshot of your wrong price is not.
Agent Capabilities
8 actionsImage Generation
2- Generate Image Generates one or more images from a text prompt with the Ideogram v3 model, the current best in class for accurate in-image text. Supports magic prompt enhancement, negative prompts, aspect ratio or exact resolution, style presets, custom color palettes, style codes, and style reference images. Every result is saved to FlowRunner file storage and returned as a durable URL with per-image metadata including prompt, resolution, seed, and safety flags.
- Generate Image (Legacy) Generates images through Ideogram's legacy endpoint for older model versions from V_1 through V_2A_TURBO. Kept for backward compatibility and for flows pinned to a specific legacy model; new work should use Generate Image. Results are saved to durable file URLs the same way.
Image Editing
4- Edit Image Inpaints part of an existing image: provide the source, a matching black-and-white mask, and a prompt for what fills the masked region while the rest is preserved. The tool for swapping a headline, fixing a detail, or dropping text into a specific area of finished artwork.
- Remix Image Blends an existing image with a new prompt, with Image Weight controlling how strongly the original steers the result. Full v3 styling controls apply. Used for seasonal variants and art-directed iterations that should stay recognizably on brand.
- Reframe Image Outpaints an image to a new resolution, generating surrounding content while keeping the original subject. The step that turns one approved asset into every aspect ratio a campaign's placements demand.
- Replace Background Replaces an image's background from a prompt while preserving the foreground subject, with v3 styling controls including rendering speed and color palettes. The workhorse for normalizing product photography onto a standard backdrop.
Image Enhancement
1- Upscale Image Raises an image to higher resolution with optional prompt guidance plus resemblance and detail controls that trade fidelity against added detail. The last step before an approved visual goes to production.
Image Understanding
1- Describe Image Returns generated text descriptions of an image's content. Doubles as automated QA on rendered text, a source for accessibility captions, and a way to reverse-engineer a prompt from a reference image. Text only, so no file is saved.
Frequently Asked Questions
What can FlowRunner do with Ideogram?
FlowRunner agents can run Generate Image, Edit Image, and Remix Image in Ideogram, plus 5 more actions.
Does connecting Ideogram to FlowRunner require OAuth?
No. Ideogram connects to FlowRunner with an API key, no OAuth flow required.
Can Ideogram trigger a FlowRunner workflow automatically?
Ideogram doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.
Start building with Ideogram
$100 in credits. No card required. Connect in minutes.