Stability AI
AIGenerate and edit images with Stability AI's Stable Diffusion and Stable Image models. Agents create images from prompts, inpaint and outpaint, replace backgrounds, upscale to print resolution, animate stills into video, and produce 3D assets and audio, with every asset saved to FlowRunner file storage.
What This Integration Enables
Stability AI is the image engine for teams that treat product imagery as inventory: hundreds of SKUs, one visual standard, and no appetite for a studio day every time something changes. FlowRunner agents generate with Stable Image Ultra, Core, and the Stable Diffusion 3 family, then work photos the way an editor would: Erase blemishes, Inpaint regions, Outpaint beyond the frame, swap objects with Search and Replace, recolor with Search and Recolor, and standardize with Remove Background and Replace Background and Relight. The Control actions turn sketches, structures, and reference styles into finished artwork; the upscalers take images to print resolution; Image to Video animates stills; Image to 3D produces GLB assets; Text to Audio and Audio to Audio cover music and sound. Every result is saved to FlowRunner file storage with a durable URL, because Stability's own result data is transient. The one call agents never make alone is the truthfulness call. An edited product photo that no longer matches the physical product is a customer-facing claim, and claims get [human-in-the-loop](/concepts/human-in-the-loop/) review before they reach a listing.
Without FlowRunner
With FlowRunner
Use Case Scenarios
Catalog standardization at SKU scale
New product photos arrive from suppliers in every imaginable style. The agent runs Remove Background on each, applies Replace Background and Relight against the brand's standard scene so lighting matches across the catalog, and upscales anything below listing resolution with Upscale (Conservative), which preserves the product faithfully instead of inventing detail. The merchandising lead approves the treated set, and approved images replace the listing photos in [Shopify](/integrations/shopify). The catalog looks like one brand shot it, because functionally one pipeline did.
Publish-ready editorial imagery
A content workflow needs a hero image, not a stock-library compromise. The agent generates candidates with Generate Image (Ultra), whose prompt adherence and typography handling suit editorial work, and on selection uploads the image to [WordPress](/integrations/wordpress) and attaches it to the draft post. For recurring formats, Control Style keeps every generated hero in the publication's established visual language, so the archive stays coherent as volume grows.
From sketch to campaign motion
A designer's rough sketch becomes finished artwork with Control Sketch, guided by the campaign prompt; Control Style Transfer carries the approved aesthetic onto new compositions. The strongest still is animated into a short clip with Image to Video, and the agent posts the result to [Slack](/integrations/slack) for creative review before it enters the campaign kit. Concepts move from whiteboard to reviewable motion inside a day, with the team judging real outputs instead of describing hypothetical ones.
Human-in-Loop Highlight
Search and Replace and Inpaint can alter the product itself, not just the scenery around it, and a listing photo that misrepresents what the buyer receives turns into returns, disputes, and marketplace takedowns. The edit costs a credit; publishing it can cost the listing. So when an edit's target region touches the product rather than the background, the agent stops and posts the before-and-after pair to the merchandising lead: "Inpaint removed the cable clutter on SKU 4417, but the edit region overlapped the device's rear panel. Compare the port layout. Publish this image to the listing, or keep the original?" Background standardization flows through on approval of the batch; anything that changes the product's appearance waits for a named yes on that specific image.
Agent Capabilities
28 actionsImage Generation
5- Generate Image (Ultra) Generates a high-quality image with Stable Image Ultra, tuned for prompt adherence, typography, and photorealistic detail, with aspect ratio, negative prompt, image reference, style preset, and seed controls. Saved to FlowRunner file storage with a durable URL.
- Generate Image (Core) Fast, high-quality general-purpose generation that needs no prompt engineering.
- Generate Image (Stable Diffusion 3) Generates with the SD 3.x family in text-to-image or image-to-image mode with adjustable strength.
- Generate Image (v1 Text to Image) Legacy v1 generation with weighted prompts and classic diffusion controls, up to 10 samples per request. Prefer the v2beta actions for new work.
- Generate Image (v1 Image to Image) Legacy v1 image-to-image with weighted prompts and strength control. Prefer the v2beta SD3 action for new work.
Editing
7- Erase Removes unwanted objects or blemishes using a mask, explicit or derived from the alpha channel.
- Inpaint Fills or replaces a masked region from a text prompt, blending seamlessly. Edits touching the product region are routed for human review in FlowRunner workflows.
- Outpaint Extends an image in chosen directions, generating new content beyond the original borders.
- Search and Replace Swaps an object without a manual mask: describe what to find and what replaces it. Routed for review when applied to product imagery.
- Search and Recolor Changes a described object's color or texture without a manual mask.
- Remove Background Segments the subject and returns a cutout with transparency.
- Replace Background and Relight Replaces the background and relights the scene to match, from a prompt or reference image, with lighting direction and strength controls. Asynchronous, with polling built in.
Upscaling
3- Upscale (Fast) Quick 4x upscaling with minimal latency and no prompt.
- Upscale (Conservative) Upscales to roughly 4 megapixels while faithfully preserving content, the choice for photography where fidelity matters.
- Upscale (Creative) Upscales degraded or low-resolution images with creative reinterpretation guided by a prompt. Asynchronous.
Control
4- Control Sketch Renders a sketch or line drawing into a finished image guided by a prompt, with adjustable control strength.
- Control Structure Preserves an input image's composition and structure while re-rendering its content from a prompt.
- Control Style Generates new images in the visual style of a reference image.
- Control Style Transfer Transfers a style image's aesthetic onto a separate content image's composition.
Video and 3D
3- Image to Video Animates a still into a short clip with Stable Video Diffusion, with motion amount and adherence controls. Asynchronous, with polling built in.
- Get Video Result Fetches an Image to Video job's result by generation ID; finished MP4s save to durable storage. Results are retained for 24 hours.
- Image to 3D Converts a single image into a textured GLB mesh with Stable Fast 3D.
Audio
2- Text to Audio Generates music or sound effects from a prompt with Stable Audio 2.5, up to about 3 minutes of stereo audio.
- Audio to Audio Transforms an existing clip according to a prompt, restyling a source track while keeping its character.
Results and Account
4- Get Generation Result Fetches any asynchronous v2beta generation by ID; finished images save to durable storage. Results are retained for 24 hours.
- List Engines Lists the generation engines available to the account, for the legacy v1 actions.
- Get Account Retrieves the account profile and organization memberships behind the API key.
- Get Balance Retrieves the remaining credit balance. The pre-flight check before every batch.
Frequently Asked Questions
What can FlowRunner do with Stability AI?
FlowRunner agents can run Generate Image (Ultra), Generate Image (Core), and Generate Image (Stable Diffusion 3) in Stability AI, plus 25 more actions.
Does connecting Stability AI to FlowRunner require OAuth?
No. Stability AI connects to FlowRunner with an API key, no OAuth flow required.
Can Stability AI trigger a FlowRunner workflow automatically?
Stability AI doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.
Start building with Stability AI
$100 in credits. No card required. Connect in minutes.