Runway Gen-4: References, Video, Pricing and Model Comparisons
Runway Gen-4 is an image-to-video model built around controllable motion and visual consistency. This guide separates Gen-4 Video from Gen-4 Image References, explains the real credit costs, and shows when Gen-4 Turbo or Gen-4.5 is the better choice.
Runway Gen-4 Video Generator
Runway Gen-4 at a glance
Required input
Gen-4 Video starts from an image plus an optional motion prompt; it is not a text-only video model.
Duration and output
Official outputs are 5 or 10 seconds, 720p, at 24 fps in six supported aspect ratios.
Official credit rate
Standard Gen-4 costs 12 Runway credits per generated second; Turbo costs 5 credits per second.
Best fit
Use it when you already have a strong first frame and need controlled camera or subject motion.
Model basics
What Is Runway Gen-4 Video?
Runway introduced Gen-4 in March 2025 as a family of media-generation models focused on consistent subjects, objects, locations and styles. The name covers connected but distinct workflows. Gen-4 Video animates a required first-frame image. Gen-4 Image creates stills, while Gen-4 References uses one or more images to carry a character, object, place or style into new stills. Those images can then become first frames for video. Treating every feature as one video-mode switch is the most common source of confusion.
In Gen-4 Video, the image supplies composition, appearance, lighting and art direction. The prompt should concentrate on what changes over time: subject action, environmental motion, camera movement, timing and direction. A line such as “the camera slowly arcs left as wind moves the grass” is more useful than redescribing every visible object. Both standard and Turbo accept an image and prompt, produce 5- or 10-second clips, and support 16:9, 9:16, 1:1, 4:3, 3:4 and 21:9 outputs at 720p.
Gen-4 is no longer Runway's newest general model—Gen-4.5 supports both text-to-video and image-to-video—but it remains practical when a team already has a first-frame workflow or wants lower-cost Turbo iteration. Choose by input type, required control and iteration budget, not the version number alone.
| Specification | Gen-4 Video |
|---|---|
| Input | Image required; text motion prompt supported |
| Duration | 5 or 10 seconds |
| Resolution | 720p |
| Frame rate | 24 fps |
| Aspect ratios | 16:9, 9:16, 1:1, 4:3, 3:4 and 21:9 |
| Official web cost | 12 credits per second |
The total is 60 credits for 5 seconds or 120 credits for 10 seconds on standard Gen-4.
Consistency workflow
Runway Gen-4 References: What It Does and How to Use It
Gen-4 References is primarily an image-generation workflow. It does not mean a Gen-4 Video request can directly juggle several reference images while generating motion. Instead, use up to three active references to build a consistent still, then send a chosen output into a video model as the first frame. This separates identity and art direction from motion design.
A reference can represent a person, product, location, object or style. After saving and naming it, call it inside a conversational prompt and combine it with other references. One image can anchor a character while another defines a room. The resulting still can place that person in the room while retaining recognizable traits. References also help create close-ups, wide shots, alternate angles and prop details that belong to the same world.
Start with a sharp image in even natural light. A neutral expression gives the model fewer accidental cues to preserve. Runway's developer guidance recommends not supplying references below 640 by 640 pixels or above 4K. Name each one clearly, keep the first prompt simple, and add one constraint at a time. With multiple sources, state which image controls identity, environment and style instead of making the model infer the hierarchy.
Once the image set is coherent, choose the strongest composition as the video input. Do not ask the motion prompt to repair identity, wardrobe, set design and movement at once. Describe movement only, generate a short draft, and review the first and last frames for drift.
Character reference
Use an evenly lit, unobstructed portrait or full-body image. Generate the required angles as stills before animating them.
Location reference
Preserve architecture, palette and atmosphere across establishing shots, inserts and alternate viewpoints.
Object reference
Keep silhouette, materials and identifying details stable while changing context or camera coverage.
Style reference
Carry lighting, texture and visual treatment without mixing too many competing style anchors.
Real capabilities
Runway Gen-4 Video Features—and Their Boundaries
Gen-4's strongest use case is image-to-video. A prepared keyframe gives a precise visual target while the prompt controls camera and scene motion. It fits product shots, atmospheric b-roll, character moments, establishing shots and short effects plates. Five or ten seconds is enough for one staged action, not a long sequence with several cuts, dialogue exchanges and persistent causal logic.
Illustrated, 3D or graphic source images can be animated through the same workflow. A restrained camera move and one subject action usually preserve the design better than many simultaneous events. For consistent multi-shot work, create a reference-based image set first and animate each shot separately.
Gen-4 Image and References handle stills; they are not hidden text-to-video settings. Video editing, video-to-video and lip sync belong to other tools. A completed clip can continue into retiming, upscaling, Act-Two or Edit Studio, but Gen-4 Video itself is not a full editor or automatic dialogue system.
Image to video
The input image fixes the visual starting point; the prompt describes motion and camera behavior.
World consistency
Build consistent stills with References, then animate each selected frame as a separate shot.
Camera choreography
Use concise terms such as push-in, arc, pan, handheld follow or locked-off shot.
Post-generation workflow
Use dedicated tools for retiming, 4K upscale, performance transfer, editing and lip sync.
Faster iteration
Runway Gen-4 Turbo vs Standard Gen-4
Gen-4 Turbo is the lower-cost, speed-oriented variant of the same image-to-video workflow. Both require a first-frame image and deliver the same published 720p aspect-ratio set. The reliable quantitative difference is cost: Turbo uses 5 credits per output second, while standard Gen-4 uses 12. A 10-second Turbo clip costs 50 credits instead of 120.
Runway recommends exploring in Turbo and switching to standard Gen-4 when a draft misses the target. That is more useful than assuming a universal speed multiplier. Queue time varies, and perceived quality depends on the source and motion. Use Turbo for prompt exploration and broad variations; reserve standard for a shot where a Turbo draft shows a specific problem worth paying to improve.
| Decision point | Gen-4 | Gen-4 Turbo |
|---|---|---|
| Credit rate | 12 credits/second | 5 credits/second |
| 5-second clip | 60 credits | 25 credits |
| 10-second clip | 120 credits | 50 credits |
| Published output | 720p, 24 fps | 720p, 24 fps |
| Best use | Selective final attempts | Drafts and iteration |
Choose the right model
Runway Gen-4 vs Gen-3 Alpha vs Gen-4.5
The models overlap, but their inputs differ. Gen-3 Alpha offered text-to-video and image-to-video. Gen-4 Video narrowed the task around a required first frame and controllability. Gen-4.5 expands again: it supports text or image input, accepts 2–10 second durations and is Runway's current quality-focused general model.
Choose Gen-4 when your project already has designed frames. Choose Turbo when iteration budget matters most. Choose Gen-4.5 for text-only generation, flexible duration, stronger prompt adherence or newer professional output options. Gen-3 Alpha may still matter to an established legacy workflow, but new projects should compare current models first.
| Feature | Gen-3 Alpha | Gen-4 | Gen-4.5 |
|---|---|---|---|
| Primary inputs | Text or image | Image required | Text or image |
| Duration | Legacy workflow | 5 or 10 seconds | 2–10 seconds |
| Best reason | Existing pipeline | First-frame control | Current quality and flexibility |
| Current position | Older generation | Older but practical | Current advanced model |
Budget and access
Runway Gen-4 Pricing, Free Access and API
The web app charges Gen-4 at 12 credits per generated second and Turbo at 5; Gen-4.5 currently uses 12. A subscription gives a monthly web-credit pool, so calculate how many drafts and final attempts a shot list can support. Plan prices and included tools can change; confirm checkout before paying.
A new Free workspace receives 125 one-time credits, but Runway exposes only a limited, changing model set. The credits do not refresh, and the live selector is the authoritative access check. A balance does not guarantee Gen-4 if the model asks for an upgrade.
Runway Dev billing is separate. Developer credits cost one cent each, and the public catalog exposes gen4_turbo at 5 API credits per second. Standard Gen-4 is not listed as a separate current public API video model; Gen-4.5 is available at 12. Select the documented identifier rather than assuming every web model has a matching endpoint.
For an economical workflow, approve the first frame before generating, use Turbo for motion drafts, and rerun only when you can name the problem being corrected. One ten-second standard attempt costs more than two ten-second Turbo attempts.
| Access route | Model | Published rate | Important detail |
|---|---|---|---|
| Runway web | Gen-4 | 12 credits/second | Requires image input |
| Runway web | Gen-4 Turbo | 5 credits/second | Lower-cost drafts |
| Runway Dev API | gen4_turbo | 5 credits/second | Separate API balance |
| Runway Dev API | gen4.5 | 12 credits/second | Text and image modes |
Rates reflect current official documentation; access and catalogs can change.
Direct answers
Runway Gen-4 Frequently Asked Questions
Continue the Gen-4 cluster
Compare the rest of the Runway Gen-4 family
Use the focused guides for a deeper speed-and-cost decision or for the newer Gen-4.5 workflow.