Name: Gen-4 References
Uploaded: 2025-05-08T17:52:18.449Z
Description: Learn how to use References in Gen-4 for consistent characters, locations and more.

Learn the basics of using Gen-1 in Runway.

- Head into AI Tools in Runway to find Gen-1
- Select an asset
- Three style reference options:
    - Image (choose your own image)
    - Presets by Runway
    - Prompt (enter a text prompt)
- Keep settings simple or dive into advanced settings: https://help.runwayml.com/hc/en-us/articles/15161225169171-Gen-1-Gen-2-Advanced-Settings
- Find recommendations for settings by hovering over tool tips


Gen-1 Basics

Gen-1

Use this quick tutorial to learn how to use Runway's mobile app for iOS.

- Download the Runway app from the App Store app
- Sign into Runway using the same account you use on the web
- Circular purple button will initiate a new generation
- 1 second of video generation = 14 credits
    - To top up credits, click profile icon in the top right corner of the Home page and tap “Credits”
- To create a video, simply hold down the record button and let go when you’re finished
- Three main style settings:
    - Presets by Runway
    - Reference image: click “Add” to choose an image from your library; adjust style strength and seed number here
    - Text prompt
- Hit “Generate” — video will be queued and show a percentage status circle
- Generate new options in the meantime; videos will continue generating in the background
- Generations will be stored in library tab


Using the Mobile App

Runway Mobile

Transform a low resolution image into a high resolution image with Upscale Image.

- Use Upscale Image to add more detail to increase image resolution
- Upload an image asset
- Select desired size under “Scale” and click “Process”
- That’s all there is to it!
- Download the image or add to your assets by hovering over it and clicking the icons on the top right


Upscale Images

Learn how to use AI Tool Erase and Replace to easily modify images.

- Head to AI Tools to find Erase and Replace
- Choose an image asset from your computer or asset library
- Adjust brush size using the slider on the right sidebar
- Brush over the portion of the image you’d like to replace
- Enter a text prompt and click “Replace”
- “Add to assets” to add to your Runway asset library
- “Download” to download locally to your computer
- Use “Clear” to generate something new with a new prompt


Erase and Replace

Learn to create a bespoke image sequence using Image to Image.

- Remove the background of your video with Green Screen and save as a PNG sequence for the most flexibility down the road
- Open Image to Image and use first frame in the sequence as a test frame
    - Describe the entire scene, touch on style that you want
    - Make sure the look is right — you’ll be committing to it in this workflow
    - Lock seed number and don’t change the prompt for images you like
    - Save to assets to make sure you can find them later
- Select every few frames if you don’t want to select them all
- String the frames together in the video editor


Image to Image

Use this tutorial to learn how to use the Text to Image AI Tool in Runway.

- Head to AI Tools to find Text to Image
- Enter a simple text prompt
- Open up “Advanced” under “Settings” sidebar to explore refinement options
    - Size, style, medium, mood, prompt weight
- Prompt weight will determine how much the text prompt is taken into consideration; higher number = most consideration, lower number = least
- Experiment with more complex prompts
- “Add to assets” to save to your Runway asset library
- “Download” to download locally to your computer


Text to Image

Bring still images to life and learn how to create a living portrait in Runway.

Step 1: Brush out and replace the parts of your image you'd like to change using either Erase & Replace or Infinite Image.
Step 2: Head over to Frame Interpolation to animate your various images into a video.
Step 3: Behold the beauty of your new living image.


Create a Living Portrait

Learn how to seamlessly transition images into one another by using our Frame Interpolation AI Tool.

- Frame Interpolation works well to create videos with subtle movement from still images
- The more frames you have to work with, the better
- Keep the differences between photos subtle if possible for realistic effect
- Try using Erase and Replace to create new images to interpolate between!


Frame Interpolation

Learn to use the Backdrop Remix AI Tool, and remix your images into anything you want them to be.

- Backdrop Remix changes the background of your photos while keeping the subject in tact
- Upload your photo, and choose a preset or a custom prompt
- Click “Generate,” and Runway will get to work replacing the background automatically — that’s it!


Backdrop Remix

Use this lesson to learn tips and tricks for using Text to Color Grade.

- The main thing to keep in mind: you can never go wrong with natural language.
- Typing in a movie title alone might be counterproductive sometimes; add in a description to make the movie title interpretation less literal.
- Same goes for specific film stocks — add additional descriptors.
- Try different seed numbers for different options.


Text to Color Grade

- Find Infinite Image in the AI Tools section. Start with a text prompt to generate an original image, or upload your own image.
- Move the generation frame to desired area. Make sure it’s overlapping the original image!
- Be descriptive with your text prompt and click “Generate.” You’ll then be presented with 4 initial generated options to choose from.
- Repeat this process as many times as you’d like to continue extending your image.
- Use adjustment button to adjust prompt weight; lower for more surprising and creative results, higher to stay more true to the prompt.
- Try opening your final image in Erase and Replace to clean up any elements in your image you’d like to change.


Infinite Image

Learn how to create generative images using various AI Tools.

- Learn how to create a generative image using AI Tools
- AI Tools used:
    - Text to Image: creates an original generated image from a text-based prompt
    - Erase and Replace: eliminates a specific part of an image and replaces it seamlessly with a new image generated from a new text based prompt
    - Image to Image: uses an image as the input, and generates a brand new image based on both the original and a descriptive text based prompt
- When writing prompts, be descriptive. Include object colors, orientations, and artistic styles in your prompt.
- Refine as you go — the initial output does not have to be perfect. You can slowly work and refine your image with tools like Erase and Replace and Image to Image to gradually get your content to where you want it.


Generative Image Workflow

AI Tools

Learn about workspaces and how they function within Runway.

- When you first sign up for a Runway account, you’ll be assigned a personal workspace, which can be viewed by clicking on the top left icon on the Dashboard
- Invite other members to your workspace to turn it into a team workspace
    - Note: you will be charged on a per person basis for each additional member
- If you’ve been invited to and accepted to a team, you can toggle between teams/workspaces using the top left icon on the Dashboard
- Make sure you are using the correct team workspace since each workspace’s plans and credits may differ


Navigating Workspaces

Learn about credits and how they’re used in Runway.

- Runway uses a credit system for generative tools and custom model training
- Your plan tier will determine how many credits are in your account
- To see plan tier and credit balance, head to “Workspace Settings” > “Plans & Billing”
- Only account owners and admins can see these settings
- 1 credit = $.01
    - Image generation: 5 credits per image
    - Gen-1: 14 credits per second
    - Gen-2: 5 credits per second
    - Custom model: 1000 credits per model
- For individual plans, credit balance will refresh every month
    - These allotments do not carry over into the following month(s)
- Additional credits may be purchased at any time
    - Purchased credits do carry over into the following month(s)
- For Enterprise plans, credit balance will refresh every month, but everyone on the team is taking credits from the same pool
    - Each Enterprise team has their own unique credit allotment
    - More credits may also be purchased on Enterprise plans if needed


Generation Credits

Learn how to add, organize and share assets in Runway.

- Navigate to “Assets” and click “Upload”
- Media is labelled by type for easy sorting later
- Folders will automatically be created after using certain AI Tools, like Gen-1 or Gen-2
- View all recent exports, or toggle between “Private” and “Shared” at the top of your Assets screen
- If you’d like to share assets with other members of your team, select the asset(s), right click, and “Move to Team”
- You can always move assets back to “Private”
- To organize assets, you can sort by many different types
- Hold down Shift + click for batch actions, like organizing or deleting


Asset Management

Learn how to navigate the Runway Dashboard.

- From the main dashboard homepage, you have access to featured AI Tools, popular AI Tools, tutorials, and community content
- If you click on “view all tools” button, you’ll see the full list of the AI Tools
- Or, you can use the categorized buttons on the left side
- “Projects” is where you’ll find all of your saved Runway projects, including editing timelines, Green Screen clips, and Inpainting clips
- “Shared with me” contains projects that you’ve been specifically invited to
- “Assets” is where all of your uploaded and creative assets will be stored
    - Sort and organize assets with folders and categories
    - Sort assets by type for easier navigation
    - Shift + click to batch organize or delete assets
    - By default, all assets that you create or upload will appear in your “Private” assets
- Dashboard may change slightly over time


Dashboard Overview

- Click on the “Invite Collaborators” button from the primary account holder’s dashboard
- Enter the emails of the team members you’d like to invite
- Choose from view access, editing access, or admin access
    - can view: can view and download assets, but can’t utilize AI Tools
    - can edit: can access all AI Tools, use credits, and manage projects and assets
    - admin: can adjust workspace roles, adjust billing information, and edit/view assets
- Typical team: 1-2 admins, plus editors
- To change name/profile image, head to “My Workspace Settings”


How to Create a Team

Learn how to join a team/workspace in Runway.

- Once you’ve been invited to a team workspace, you will get an invite to your email inbox
- Log in using Google, Apple, or sign up with Runway
- Team workspace is indicated by top left name and icon on your dashboard


How to Join a Team

Getting Started

Learn how to use Text to Image for use with Gen-2 in Runway.

- Open Text to Image and explore the newly updated styles in the “Advanced” menu
- Customize output with mediums and moods
- A notable mention is the “Cinematic” style — adept at crafting photorealistic, cinematic visuals
- Once you have a generation you like, navigate over to Gen-2, and use the image as your prompt
- Before generating, open the advanced camera controls to set the direction and intensity of motion in your video


Image to Video Workflow

Gen-2

Learn how to use Generative Audio to add speech to your images and videos.

- Find Generative Audio on the Runway dashboard
- At the top, you’ll find an area to write anything you want and turn it into a spoken audio file
- After you type out your text, you can preview and choose voices from the default voice menu
- Once you choose a voice, click on the “Generate” button
- Generate times depend on script length, but are generally pretty quick
- Audio generations are automatically saved to the “Generative Audio” folder inside of the main assets folder, but can be saved elsewhere
- Custom Voice training only requires a few minutes of clean audio with no background noise — import or record in Runway
- Finally, to create a Lip Sync video, you’ll need a video or image of a person whose face is fully in the frame
    - upload your own or use one of the characters in the list
    - Lip Sync can be used with generated audio from Text to Speech, recorded audio or uploaded audio
    - choose a voice and click “Generate”
- Tip: turn your image into a video with Gen-2, then upload that video into the Generative Audio tool to add Lip Sync to your video
- If your audio is longer than your video, once your video reaches the end of its duration, it will reverse and go back to the beginning for the duration of the audio
    - avoid using general camera motion controls and instead use Motion Brush, which makes the reverse effect less noticeable


Generative Audio Overview

Audio

Learn how to generate using Styles in Frames.

- Enter a Generative Session, then click on the paint palette icon to access Styles
- Click on any Style to activate it, then enter a prompt into the prompt box
- Simple and long prompts both work well with Styles, with primary focus on the subject and environment
- Choose an aspect ratio, then click “Generate”
- To see the prompt that was used for any of the Styles, click on the top right corner of any of the Styles thumbnails
    - This is helpful to learn what prompt modifiers are being used to create the Style
- To create your own Style, click on “Create,” give it a name, and write the prompt in the prompt box
    - Creating your own Style is a fast and easy way to create images that are in the same stylistic universe, without the need to type the same prompt structure over and over again
    - Focus on stylistic elements (color tones, lighting, shadows, textures, etc.) and avoid things that are too specific, like subject or location details to ensure your Style works for a wide variety of prompts
    - For specific location Styles, be as descriptive and detailed as possible
- Try using Aesthetic Range with Styles to continue exploring new ideas and environments


Using Styles in Frames

Learn how to generate images with Frames.

- Open a Generative Session and select the “Image” tab
- Choose aspect ratio, then start prompting
- To get the most out of Frames, it’s helpful to structure your prompt with the following format: [art style] [subject] [scene] [lighting] [color]
    - For example: A cinematic photo of a peacock in the water. The lighting is even, creating subtle shadows, with a muted natural color palette.
- Within each section of the prompt, you can keep descriptions general or choose to get more granular for more specific results
- It’s not necessary to include all of these categories every time - Frames does well understanding simple prompts as well
- Experiment with Aesthetic Range, which allows you to dial up and down the overall creative variation that you’ll see across a set of four generations
    - Lower values will remain stylistically close to the original prompt, with minimal changes to camera angles, lighting and color grading
    - Higher values will introduce more extreme stylistic changes to your image while keeping your subject the same
- Once you’ve generated images, you can download, favorite or click the “vary” button to generate four more similar images
- Click the “Use” button to swap the Generative Session UI into video mode, using the Frames output as an Image to Video input
- Check out [runwayml.com/worlds-of-frames](http://runwayml.com/worlds-of-frames) for inspiration


Generating with Frames

Frames

Learn how to use a restyled first frame for more control with Video to Video.

- Inside of a Generative Session, choose a video you’d like to stylize
- Click on “Styled first frame” icon
    - From here, you can download the first frame of your original video to apply stylization to it
    - During the stylization process, make sure the image dimensions remain the same
- Upload the styled frame into Video to Video
- Add a text prompt for additional context if you want, then click “Generate”
- Default Structure Transformation is set to 3, but try experimenting with different values to get different video motion results
- You can even try adding or changing elements from the original frame instead of completely restyling it, to turn Video to Video into a VFX engine
- If you feel like getting experimental, you can upload a completely different image than the first frame, which will give you unexpected and interesting results
    - For this method, increasing the Structure Transformation above 5 is typically helpful for getting better looking videos


Using Restyled First Frame

Learn how to use Character Reference Videos in Act-One.

- Head to Generative Sessions tool and open up Act-One in Gen-3 Alpha or Gen-3 Alpha Turbo
- You can now use a video as your character input, in addition to images
- Upload or record your driving performance, choose a video clip for your character reference video, adjust motion intensity if you’d like, and hit “Generate”
- Keep in mind: if your driving performance video is longer than your character reference video, the output will automatically reverse at the end of the clip to fill the duration
- Character reference videos in Act-One can be live action videos, giving you the freedom and flexibility of changing performances in post
- You can also bring in your Gen-3 Alpha videos as character videos, giving your generative workflows a huge boost in expressiveness
- You can even run your video clips through Video to Video before bringing them into Act-One to completely control every aspect of your performance


Character Videos in Act-One

Learn how to use Expand Video in Gen-3 Alpha Turbo.

- Expand Video allows you to seamlessly generate beyond the edges of a frame to turn horizontal videos into vertical videos, and vice versa - without compromising details or framing
- Open Gen-3 Alpha Turbo to use Expand Video
- If you upload a horizontal video, the model will automatically default into turning it into a vertical video
- You’ll have a few options:
    - Click the Generate button without providing a text prompt
    - Provide a detailed text prompt telling the model what you’re hoping to see in the parts of the frame that aren’t visible
    - Provide the model with an outpainted image of the first frame of your video
- Best practices:
    - Subject is close to center of the frame
    - No graphics or text present in the video
    - Use a guiding text prompt
- If you run multiple expansions in a row, you can create wide shots of the scene — then by overlaying the original footage onto the expanded canvas, you can create super zooms


Gen-3 Alpha Expand Video

Learn how to use Camera Control in Gen-3 Alpha Turbo.

- Head to Generative Video, then Gen-3 Alpha Turbo and upload an image
- Camera Control allows you to plan out specific camera movements for your scene: left/right, up/down, in/out, pan left/right, tilt up/down, rotate clockwise/counter-clockwise
- The camera speed is determined by how close your subject is to the camera
    - Use smaller values when your subject is closer and larger values when your subject is far away
- Camera Control will generally work better if you provide a text prompt to help guide your generation (example: where you want the camera to move, what you want your subject to be doing, add more to the scene)
- Try using presets for inspiration, found underneath the prompt box


Gen-3 Alpha Camera Control

Learn how to use Act-One to generate expressive performances using a single driving video and character image.

- Head into Generative Video with Gen-3 Alpha selected
- Click “Act-One” icon on the left sidebar
- Add your source footage
    - Act-One can work with characters positioned at varying angles as long as it’s not too extreme
    - Maintain lively facial expressions with the entire face in view, with minimal body movement and clear audio
- Add your reference image
    - Upload your own image or choose from one of our presets
- Once ready, hit “Generate”
    - Your result will be a seamless animation preserving all of the subtle expressions and movements from your performance, but transposed onto the character from your reference image


Getting Started with Act-One

Learn how to use Video to Video in Gen-3 Alpha.

- Open up Generative Video and select Gen-3 Alpha
- Upload a source video
    - input videos 5s or less cost 50 credits, input videos longer than 5s (up to 10s) cost 100 credits
- Add a text prompt describing how you’d like to stylize your video
- Adjust structure transformation slider in settings
    - lower values will adhere to the source video structure more
    - higher values will allow you to slightly deviate from structure of input video
- Click “Generate”
- Use custom style presets to experiment with new styles
    - edit the purple text in the text prompt with a quick description of your scene


Gen-3 Alpha Video to Video

Learn how to extend Gen-3 Alpha generations.

- Choose Gen-3 Alpha from the model dropdown in the Text/Image to Video tool (extensions aren’t available in Gen-3 Alpha Turbo yet)
- Generate a video from a text prompt or an image
- Click “Extend” — choose to extend it by 5s or 10s for a maximum of three extensions total
- If you click the “Extend” button with no additional prompt, Gen-3 Alpha will automatically continue the motion from the previous video
- If you’d like, you can add a text prompt to your extension for more control, or to add something new
- From your assets folder, you can always navigate back to a previous Gen-3 Alpha generation and extend it from there
- Extensions are a great way to get more out of Gen-3 Alpha — showcasing the environment, transitioning the scene or just continuing the action


Gen-3 Alpha Extensions

Learn how to blend live action video with Generative Visual Effects.

- Find or shoot a source video
    - Suggested camera settings: auto focus off, auto white balance off, camera on tripod
- Next, scrub through your footage and export a still image from where you want to add visual effects
- Bring the still image into Gen-3 Alpha Image to Video
- Depending on your idea, you can choose to input the image as the first frame or last frame
- In this example, we’ll select the first frame and add a text prompt describing the effect
- Once you have a shot you like, composite (combine two or more images together to make one final image) the source video with the generated video
- For the final output, feel free to add color correction or masking if needed
- Gen-3 Alpha Image to Video allows you to achieve various levels of VFX — from singular assets to fully blended live action and generated content


Generative Visual Effects

Learn how to use Image to Video with Runway’s newest video model, Gen-3 Alpha.

- From your Runway dashboard, click on “Text/Image to Video” and upload an image
- If you’d like, you can click the “Generate” button with no additional prompt guidance, and the model will interpret the image to give you the best results
- You can also provide a text prompt alongside your image which will help guide the result better or, introduce new elements or movements throughout the video
- Image to Video is a great way to start with a specific style, character or composition for added control and intention when generating


Using Image to Video in Gen-3 Alpha

Learn how to use Text to Video with Runway’s newest video model, Gen-3 Alpha.

- Head to Text/Image to Video tool and select Gen-3 Alpha from the model dropdown
- Add a text prompt
- Select 5s or 10s video to generate
- 720p generations take about 60s for a 5s video, or 90s for a 10s video
- The more detailed you can make the text prompt, the more Gen-3 Alpha will shine
    - Include things like subject, scene, lighting, camera, movements and more
- In addition to visual details, Gen-3 Alpha can be prompted for subject action, camera action, speed, transition and more
- Gen-3 Alpha works with a wide variety of prompt structures, from super simple to more complex
- If you’re looking for inspiration, separate prompt into a visual description and a camera description
    - Example: Visual: A pillow fort in a cozy living room. The pillow fort is made from an assortment of quilts, fabrics and pillows Camera motion: Hand held camera smoothly zooms into the entrance of the pillow fort, revealing an ancient castle in the interior
- For more helpful resources, check out our [Help Center](https://help.runwayml.com/hc/en-us) or [Discord](http://discord.gg/invite/runwayml)


Getting Started with Gen-3 Alpha

Gen-3 Alpha

Learn how to use 3D assets in Gen-4 References.

- Using 3D assets as part of your generative workflow allows you to control specific scene elements contained in your reference image
- While you can’t use 3D models directly in Runway, you can still create pre-comp images that can be brought in and used with References
- Start by using your 3D application of choice
    - Build out a very basic 3D scene
    - Use the 3D camera to get the angle that you’re looking for
    - Export the camera POV as a single still image
- Import that image into References, and combine it with a style reference image and a text prompt
    - The prompt should look something like “render image 1 in the style of image 2”
- Try swapping out image 2 with different style references for different results
- Once you get an image you like, try asking References for different angles of the scene, or closeups on specific elements
- Choose your favorite images, turn them into videos with Gen-4, then edit them together for your final video result


Using 3D Assets in Gen-4 References

Learn how to use References in Gen-4 for consistent characters, locations and more.

- Gen-4 References allows you to visualize and create consistent characters and locations, new shots and angles and so much more
- Open a new Session and click on the “Image” tab
- Drag in up to 3 Reference images to use
- If you’d like to save a Reference into your library for a quick recall later, hover over the Reference image to give it a name
    - Use Saved References by using the @ command in your text prompt
- Be conversational with your prompts - ask for new angles or characters
- Choose any part of the image to reference - lighting, color grading, compositions, poses and more
- Reference different images directly by using tags/saved References, or simply reference “image 1” or “image 2” in your prompts according to the upload order of your images
- Currently, References primarily supports character and location preservation, but we will be making updates in the future to support objects, styles and more


Gen-4 References

- Open a Generative Session and make sure that Gen-4 is selected from them model dropdown
- Upload an image to use it as the first frame of your video
- Gen-4 supports Image to Video with multiple aspect ratios and will automatically adhere to the closest aspect ratio to your image, but feel free to change it
- Providing a text prompt with your image provides more control of your scene and camera
- Prompting should focus on the things you want to happen in the video, specifically the subject action and camera action
- Or, you can use Gen-4 with no prompts at all for stunning results
- Gen-4 works well in a visual effects workflow, prompting for scene actions to take place, or by animating visual effects assets that can be used in a more traditional compositing workflow
- Don’t forget to combine Runway workflows and models together, like Act-One to make your characters speak, or Expand Video to see more of your scene, run your outputs through Video to Video to reimagine them, or upscale your videos to 4K for higher fidelity
