Runway
AIGenerate media with Runway inside your workflows. Agents create images and video from text or reference media, dub and isolate audio, build avatar and product ad videos, and run Runway workflows, then wait on tasks and save outputs to files across 51 actions.
What This Integration Enables
Runway is a full production studio behind an API: image and video generation, audio and dubbing, talking avatars, and purpose-built marketing recipes for product ads, UGC-style videos, and ad localization. Two operational facts shape how it should be run. Output URLs expire within 24 to 48 hours, and every generation draws from a fixed credit pool with per-model limits. This integration is built around both: agents generate across the entire surface, then persist results with Save Task Output to Files before links die, and check Get Organization Info before batches so spend is a decision rather than a surprise. Agents handle the throughput: rendering, dubbing, upscaling, fanning a master creative across markets. People hold the calls that multiply: which sample is right before a batch runs, which cloned voice may represent the brand, which avatar fronts the company. That split is the [human-in-the-loop](/concepts/human-in-the-loop/) discipline the rest of this page keeps returning to.
Without FlowRunner
With FlowRunner
Use Case Scenarios
One master ad, every market
The quarterly campaign master gets approved. The agent checks the balance with Get Organization Info, runs Localize Ad Image for the first target market, and posts the sample to [Slack](/integrations/slack) for a native-speaker check. On approval, it fans out across the remaining languages, persists each output with Save Task Output to Files, and archives the set to [Google Drive](/integrations/google-drive-api). What used to be a per-market vendor cycle becomes a single reviewed run.
Product UGC videos from a campaign table
The agent pulls product photos and creative briefs from [Airtable](/integrations/airtable) with Get Records, runs Create Product UGC Video for each row to produce vertical presenter-style videos, and writes the durable video URLs back to the campaign table with Create Record. Create Product Ad Video and Create Product Campaign Images cover the polished end of the same pipeline, and Swap Product in Video re-targets a proven reference video to a new SKU without a reshoot.
A branded spokesperson, built with consent and staging
The team designs the brand's spokesperson voice with Preview Voice, iterating on the description without persisting anything. Once the brand owner approves a preview, Create Voice makes it real, Create Avatar binds it to an approved reference image and knowledge documents, and Generate Avatar Video produces updates on demand. Dub Audio extends finished videos into 29 languages in the original speaker's cloned voice, with the same review gate per market before anything ships.
Human-in-Loop Highlight
Localize Ad Image will happily render a flawed master into 22 languages, and credits spent on a broken batch cannot be recalled. Worse, the output is paid creative headed for markets where often nobody on the core team reads the language. So the fan-out never runs first. The agent localizes one market, then posts: "German variant of the Q4 master attached. Current balance from Get Organization Info: enough for the full 21-market fan-out. Approve for fan-out, or fix the master first?" A regional reviewer catches the truncated headline or the wrong formality register once, before it is multiplied. The same one-then-many rule governs every batch recipe on this connector, because the unit of failure is never one image, it is one image times every market it reaches.
Agent Capabilities
51 actionsImage and Video Generation
7- Generate Image Generates images from a text prompt across Runway's model lineup, including Gen-4 Image and GPT, Gemini, and Seedream families, with reference-image guidance.
- Upscale Image Upscales an image up to 16x with the Magnific Precision Upscaler v2, with sharpening and detail controls.
- Image to Video Animates a still image into video across Gen-4.5, Gen-4 Turbo, Veo, Seedance, Gemini, and HappyHorse models, with per-model duration limits.
- Text to Video Generates video from a prompt alone, with reference media support on Seedance 2 models.
- Video to Video Transforms an existing video by instruction: restyle, add or remove elements, change environments, or convert aspect ratio.
- Upscale Video Upscales video up to 4K with creativity, sharpening, and FPS controls.
- Character Performance Animates a character image or video with the movements, expressions, and speech of a real performance video using Act-Two.
Audio
5- Generate Sound Effect Generates a sound effect from a text description, with duration and looping controls.
- Text to Speech Converts text to speech with preset voices or a voice cloned from reference audio.
- Speech to Speech Replaces the voice in a recording while preserving the original delivery and timing.
- Dub Audio Dubs a recording into another language across 29 targets, cloning the original speakers' voices by default.
- Isolate Voice Extracts clean speech from a recording, removing noise, music, and ambience.
Avatars
10- Generate Avatar Video Generates a talking avatar video from a preset or custom avatar, speaking supplied audio or text.
- List Avatars Lists the organization's custom avatars with status, voice, and personality.
- Create Avatar Creates a custom avatar from a reference image with a personality, voice, and optional knowledge documents. Asynchronous.
- Get Avatar Retrieves an avatar's processing status and configuration.
- Update Avatar Updates an avatar's image, personality, voice, or documents.
- Delete Avatar Permanently deletes a custom avatar. Routed through human approval in FlowRunner workflows.
- List Avatar Conversations Lists recorded realtime avatar sessions, filterable by avatar and date.
- Get Avatar Conversation Retrieves a session's full transcript, recording URL, and available tools.
- Delete Avatar Conversation Permanently deletes a recorded conversation, including transcript and recording. Routed through human approval.
- Get Avatar Usage Returns realtime avatar session usage over a date range with a per-day breakdown.
Voices
6- List Voices Lists the organization's custom voices with processing status and previews.
- Create Voice Creates a custom voice by cloning from a sample or designing from a description. Asynchronous.
- Get Voice Retrieves a voice's status, preview, or failure reason.
- Update Voice Updates a voice's name or description.
- Delete Voice Permanently deletes a custom voice. Routed through human approval.
- Preview Voice Generates a short preview of a designed voice without persisting it, the staging step before Create Voice.
Knowledge Documents
5- List Documents Lists the knowledge documents avatars draw on, with usage mapping.
- Create Document Creates a text document avatars can reference when answering.
- Get Document Retrieves a document's content and the avatars using it.
- Update Document Updates a document; avatars pick up changes automatically.
- Delete Document Permanently deletes a document; referencing avatars stop using it.
Marketing Recipes
7- Localize Ad Image Translates and re-renders the text in an ad image into another language while preserving layout, across 22 targets. Batches run one-sample-first with human review.
- Create Marketing Stock Image Generates marketing-ready stock images from a prompt, up to 4 variations per run.
- Create Product Ad Video Generates a product advertisement video from product photos, style images, and a creative concept.
- Create Product Campaign Images Generates campaign-style marketing images featuring a product.
- Swap Product in Video Replaces the product shown in a reference video with a different product from supplied images.
- Create Multi-Shot Video Generates a multi-shot video with automatic or custom per-shot prompts and durations.
- Create Product UGC Video Generates a vertical UGC-style video of a character presenting a product.
Tasks and Files
5- Get Task Retrieves a generation task's state, progress, and outputs.
- Wait for Task Polls a task to a terminal state, optionally copying outputs into FlowRunner file storage.
- Cancel Task Cancels a pending or running task, or deletes a finished one.
- Save Task Output to Files Persists a succeeded task's outputs to FlowRunner file storage before Runway's 24-to-48-hour URLs expire. The step no production flow skips.
- Upload File to Runway Uploads a file to Runway's ephemeral storage, returning a runway:// URI valid for 24 hours.
Workflows
4- List Workflows Lists the organization's published Runway workflows and versions.
- Get Workflow Retrieves a workflow version's description and node graph.
- Run Workflow Runs a published workflow version with optional per-node input overrides.
- Get Workflow Invocation Retrieves an invocation's state, progress, and outputs.
Organization
2- Get Organization Info Returns the credit balance, tier limits, and current usage. The pre-flight check before every batch.
- Get Credit Usage Returns day-by-day credit usage by model over up to 90 days.
Frequently Asked Questions
What can FlowRunner do with Runway?
FlowRunner agents can run Generate Image, Upscale Image, and Image to Video in Runway, plus 48 more actions.
Does connecting Runway to FlowRunner require OAuth?
No. Runway connects to FlowRunner with an API key, no OAuth flow required.
Can Runway trigger a FlowRunner workflow automatically?
Runway doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.
Start building with Runway
$100 in credits. No card required. Connect in minutes.