APIFRAME
AIReach Midjourney, Kling, Sora, Veo, Flux, Seedream, Suno, and more than 80 other image, video, and music models through the single APIFRAME v2 API. Agents generate media assets without managing an account per model.
What This Integration Enables
Anyone can generate a first image. What separates a usable creative pipeline from a demo is everything that happens next, and that is what APIFRAME actually exposes. Run Midjourney Action mirrors the buttons under a Discord grid: Upsample renders one frame of a four-image grid at full resolution from U1 to U4, Variation produces a fresh grid riffing on one frame from V1 to V4, and Inpaint, Outpaint, and Pan act on a completed upsample to repaint a region or extend the canvas. Run Suno Action does the equivalent for music, with Extend, Cover, Add Vocals, and Stems. Run Udio Action adds Extend and Remix. The creative revision cycle stops being something a person does in a chat window and becomes something a flow can run, pause inside, and record.
Underneath that sits a single job model across more than 80 image, video, and music models, including Midjourney, the Flux and Seedream families, Ideogram, Imagen, DALL-E, GPT Image, Qwen, Wan, Kling, Sora, Veo, Runway, Luma, Hailuo, Grok, Suno, Udio, Mureka, and Lyria. Everything is asynchronous: an action returns a job ID and a QUEUED status, and Get Job carries it through PROCESSING to COMPLETED or FAILED with a percentage and the credit cost. Results are CDN URLs kept for 90 days, which is generous compared with the neighbors in this category but is still a clock, so anything you intend to keep gets downloaded.
Two capabilities are easy to miss and change what you can build. Create Asset pre-uploads media into provider storage so it can be referenced later as an asset:// URI, which is how you feed large files or media behind authentication into a generation request. And the LoRA operations manage personal models trained on this account for AI Photos, which is how a recurring subject stays consistent across a campaign instead of drifting between renders.
Without FlowRunner
With FlowRunner
Use Case Scenarios
Concept to final key visual with a human in the middle
The agent generates a Midjourney grid from the approved brief and posts all four frames to the creative channel in Slack. The art director picks one. The agent upsamples that frame, runs a pan to widen the composition for the banner crop, upscales the result, and files the CDN URL against the campaign record. The machine does the six mechanical steps. The person makes the one aesthetic call, which is the only step that was ever theirs.
A campaign track written, extended, and split into stems
A launch video needs original music. The agent submits Generate Music to Suno and gets two candidate tracks back. After the team picks one, Run Suno Action extends it to the length of the cut, then runs again with Stems to split the vocal and instrumental into separate files so the edit can duck the vocals under the voiceover. All of it is job records with credit costs, so the cost of the music is a number rather than a guess.
Consistent product imagery from a trained subject
A trained LoRA holds a recurring product or spokesperson. Before generating, the agent calls Get LoRA to confirm training has finished, generates a set of images referencing that subject, removes backgrounds with Remove Image Background for the e-commerce cuts, and writes the finished assets against listings in Shopify. The subject looks the same in every render because the same trained model produced all of them.
Human-in-Loop Highlight
Delete LoRA is the one operation on this connector that an agent should never complete on its own, and the reason is precise. Removing a trained LoRA is permanent. Images already generated with it are unaffected, but every flow that still references that LoRA ID fails on its next generation request, and getting the subject back means training it again from the original source material, if that material still exists. So when a cleanup routine identifies LoRAs that have not been used in months, it does not delete them. It lists them with their last use, cross-references which live flows reference each ID, and asks the owner in Slack: "Four trained subjects are unused for 90 days. Two of them are still referenced by the seasonal campaign flow, which will fail on its next run if they go. Delete the other two, delete all four, or keep everything?" The person answers once and the agent executes exactly that. Storage tidiness is not worth a broken campaign flow discovered by an art director at 6pm, and an agent has no way to know which trained subject someone is planning to use next quarter. Deciding what stays is not an inference problem. It is somebody's call.
Agent Capabilities
19 actionsGeneration
3- Generate Image Submits an image generation job to any model APIFRAME fronts, including Midjourney, Flux, Seedream, Ideogram, Imagen, DALL-E, GPT Image, Qwen, Wan, Kling, and Grok. Returns a job ID and QUEUED immediately, so poll Get Job until COMPLETED for the CDN URLs. Model-specific options travel in Model Parameters under that model's own key.
- Generate Video Submits a video generation job to Kling, Sora, Veo, Runway, Luma, Hailuo, Seedance, Wan, Grok Imagine, or Midjourney Video. Returns a job ID to poll, and the completed result carries a CDN MP4 URL. Midjourney Video is the exception and returns four video URLs. Per-model controls such as start and end frames, resolution, or duration go in Model Parameters.
- Generate Music Submits a music generation job to Suno, Producer, Mureka, Udio, ElevenLabs Music, or Lyria. In default mode the prompt describes the music you want; in a model's custom mode the prompt is treated as lyrics. Suno, Udio, and Producer return two tracks per request while Lyria and ElevenLabs return one.
Editing and Enhancement
5- Upscale Image Enlarges an image without a prompt using Topaz Image Upscale, which preserves real photographic detail conservatively, or Clarity Upscaler, which behaves like a creative hi-res fix and suits AI art. Source URL and options travel in Model Parameters.
- Remove Image Background Cuts the subject out and returns it on a transparent background. Bria's RMBG 2.0 handles hair, fur, glass, and motion blur cleanly and suits product work, while 851-labs is the cheaper choice for high volume where edge quality matters less.
- Edit Image Repaints or extends an existing image with Flux Fill Pro. Inpaint mode uses a black-and-white mask where white repaints and black preserves; outpaint mode extends the canvas by a preset such as zoom out or a directional outpaint and ignores the mask.
- Edit Video Rewrites an existing clip from a prompt, covering outfit swaps, background replacement, and restyling such as film grain or a colour regrade, while keeping the subject's identity and motion intact. The source clip must be two to ten seconds, and a reference image can anchor the target look.
- Upscale Video Enhances an existing clip with Topaz Video Upscale, combining spatial super-resolution with frame interpolation for outputs up to 4K at 60 fps. A restoration step that takes no prompt and never changes the content of the clip.
Creative Follow-ups
3- Run Midjourney Action Runs a Midjourney follow-up on a completed job, mirroring the buttons under a Discord grid. Upsample renders one frame of a grid at full resolution from U1 to U4 and Variation produces a fresh grid from V1 to V4, both requiring an index. Inpaint, Outpaint, and Pan act on a completed upsample rather than on a grid. Credits are deducted at submission and refunded on failure.
- Run Suno Action Runs a Suno follow-up on one of the two tracks a completed Suno job produced. Extend continues the track, Cover reinterprets it in a new style, Add Vocals sings over an instrumental, and Stems splits it into separate vocal and instrumental files.
- Run Udio Action Runs a Udio follow-up on one of the two songs a completed Udio job produced. Extend continues the song with additional material and Remix reinterprets it under new guidance, and both return two new songs.
Jobs
2- Get Job Retrieves a single job by ID with its status, percentage progress, credit cost, and result. Status moves through QUEUED, PROCESSING, and then COMPLETED or FAILED. Result files sit on the APIFRAME CDN for 90 days, so download anything you need to keep.
- List Jobs Returns a cursor-paginated list of generation jobs with their statuses and results, optionally narrowed to a status or a model and scoped either to your own jobs or to every job on the team. Used for reconciliation and for reporting what a campaign actually cost.
Assets
2- Create Asset Pre-uploads a media file into provider storage from a public URL so later generation requests can reference it as an `asset://` URI instead of a direct link. This matters for large files and for media behind authentication. Upload is asynchronous, so poll Get Asset before using the ID.
- Get Asset Checks whether a pre-uploaded asset has finished processing. Status is Processing while the provider fetches and stores the file, Active once it can be referenced, or Failed with an error message.
Trained Models
3- List LoRAs Returns the personal LoRA models trained on this account for AI Photos, each with its identifier and training state. Used to pick a trained subject before generating personalized images.
- Get LoRA Retrieves a single trained LoRA with its current training status and metadata, which is how a flow confirms a subject has finished training before generating with it.
- Delete LoRA Permanently removes a trained LoRA from the account. Images already generated with it are unaffected, but any flow still referencing that LoRA ID will fail on its next generation request. Treat this as an approval-gated operation rather than a cleanup step.
Account
1- Get Account Info Returns the authenticated user, their team, and the API key used for the request, including remaining credit balance, subscription plan, and the maximum number of jobs that can run at once. Check the balance before a batch so a run does not stop part-way.
Frequently Asked Questions
What can FlowRunner do with APIFRAME?
FlowRunner agents can run Generate Image, Generate Video, and Generate Music in APIFRAME, plus 16 more actions.
Does connecting APIFRAME to FlowRunner require OAuth?
No. APIFRAME connects to FlowRunner with an API key, no OAuth flow required.
Can APIFRAME trigger a FlowRunner workflow automatically?
APIFRAME doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.
Start building with APIFRAME
$100 in credits. No card required. Connect in minutes.