Introduction
Runway Agent is an AI creative partner that works with you to produce complete, multi-shot videos through conversation. You describe what you want to make, develop the outline together, and the agent handles scene planning, shot generation, voiceover, dialogue, music, and assembly.
A built-in timeline editor lets you cut, reorder, and upload additional assets after generation.
This article walks through the basics of generating your first video with Agent. See the Spec details section for more details on outputs and credit costs.
Step 1 – Accessing Agent
Agent lives in the main left sidebar of your Runway workspace alongside other tools as a creation mode.
- Sign in to your Runway account
- In the left sidebar, click Agent
You'll land on the Agent prompt screen, where you'll describe your video and configure its settings.
Step 2 – Prompting for the video
The prompt sets the creative foundation of what Agent generates.
- Click into the Describe your video… field at the top of the panel
- Enter a description of the video you want to create
- Include the subject, setting, key actions, and any narrative arc you have in mind
Think of your prompt as a simple description of your concept. Agent will break your idea into multiple shots, so context like location, video style, and story progression all help. A single, simple sentence works well in most cases, but you can add more detail as needed.
Adding reference images
If you have specific visuals you want Agent to incorporate — characters, locations, products, or style references — you can upload them directly.
Click the + button below the prompt field or drag and drop images from your computer. Agent will use these assets when planning shots and preparing scenes.
Choosing a tone
The tone shapes the story arc, audio mood, and visual and editorial style of the final video. Under Choose a tone, click one of the preset options: Bold, Cinematic, Clean, Dramatic, Energetic, or Luxury.
If none of the presets fit, click Add your own… and describe the tone in your own words.
Step 3 – Configuring the settings
Settings control the technical output of your video. Each can be adjusted from the Settings section, and you can make final adjustments after the outline is complete.
-
Aspect ratio — Choose the frame shape (e.g.
16:9) based on where the video will be used. - Duration — Choose between 15s or 30s durations. Longer durations cost more credits.
-
Resolution — Choose a
720por1080poutput resolution. - Audio — Choose an audio mode to control how sound is generated.
Credit costs scale with the selected duration, resolution, and other settings or generations created while iterating. Review the Spec details section for more details.
If you're testing a concept and want to save credits, start with a shorter duration and lower quality before committing to higher cost settings.
Step 4 – Creating and reviewing your outline
Once your prompt, assets, tone, and settings are set, Agent will generate an outline — the shot-by-shot plan it uses to build the final video.
- Click Create my outline in the bottom-right of the panel.
- Review the generated outline.
This is your chance to confirm satisfaction with the plan Agent created before producing shots. We recommend reviewing this before starting the generation process.
If changes are needed before starting the generation, you can provide feedback directly through the chat for automated updates or manually update the summary, visual references, story beats, and settings by clicking or interacting with the settings.
Editing Visual References
Hovering over an image and clicking the ellipsis (...) button reveals more options to work with your image:
- Edit image — Hone in on this exact image for further iteration through chat
- Replace image — Replace this image by uploading a different one in its place
- Rename — Change the reference name and type (character, brand, environment, prop, or style)
- Delete — Delete the visual reference from the outline
Visual Reference names & types
When you select the Rename option for a visual reference, you'll be able to customize both the Reference Name and Type.
The Reference Name is the human-readable name used to refer to the image in the story and subsequent generations. We recommend keeping this simplistic for the best results.
The Type determines how Agent leverages the image to create the video:
- Character — Prioritizes consistency of a person or figure across scenes
- Brand — Prioritizes consistency of a product or logo across scenes
- Environment — Treats the image as a background or setting
- Prop — Treats the image as an object that may be interacted with or reused in a scene
- Style — Uses the image only as a style reference
Agent automatically sets the Type, but you can always adjust as needed.
Step 5 – Generating and touching up the video
After you approve the outline, Agent generates each shot, edits them together, and adds sound.
When the video is complete, you can cut, reorganize, and upload additional assets using the timeline editor:
Split
Split separates media files into two. To use Split to separate a single media clip:
- Click the media you want to split in the timeline editor
- Move the scrubber to the location to split
- Click Split
Selecting a media clip (which adds the white outline) will split only that file. To split all tracks at a point, make sure no single media clip is selected.
Trim a clip
Hover near the edge of the media clip and click and drag the edges to trim.
Adjust track volume
Click the track's left icon to adjust the volume. The volume adjustments apply to all clips within the track.
Add a new media track
Click the left-hand + symbol to add a new track to work with. Use this to upload new media to layer into your video.
Undo and Redo edits
Undo and Redo allow you to quickly test minor adjustments in the timeline editor without edit permanence:
- Use Undo to quickly remove the last change(s) to the timeline
- Use Redo to revert the last Undo action
If larger changes are needed to the overall story or generations, you can revert to Outline mode to further iterate on the story beats and reference images.
Examples
| Visual reference | Prompt | Mood | Output |
|---|---|---|---|
| we follow the kite as it escapes a child's hand and floats across universes | Clean |
||
claymation style product ad featuring two crabs fighting over a can of fizz. one of the crabs says "don't be so shellfish" |
Fun |
||
|
|
both characters wear 1980s dance team attire (sequins, bold colors) and sit in the front row of theatre seats near an empty stage. they talk about his lack of commitment to the dance team. | Dramatic |
Next Steps
You now know how to generate a multi-shot video with Agent — from prompt to finished cut. From here, you might want to use:
- The Upscale Video app to upscale to 4K quality
Spec details
This section outlines the available settings, output formats, and includes a table for credit cost estimation.
Setting & output details
| Detail | Value |
|---|---|
Supported inputs |
Text prompt, optional reference images or assets |
Output format |
Multi-shot video with edits and sound |
Aspect ratios |
16:9 |
Resolution |
720p |
Duration |
15s |
Audio options |
Music + Dialogue |
Credit cost |
Varies based on output duration + resolution (see below) |
Credit details
| Video | ||
|---|---|---|
| Resolution | Duration | |
| 15s | 30s | |
720p |
540 credits |
1080 credits |
1080p |
600 credits |
1200 credits |
Audio | ||
Audio selection |
Duration |
|
15s |
30s |
|
Music + Dialogue |
8 credits |
8 credits |
Music + Voiceover |
Up to 11 credits |
Up to 14 credits |
Music only |
8 credits |
8 credits |
Image | ||
Per new image created |
20 credits |
|