Runway Gen-4
Last reviewed
May 17, 2026
Sources
47 citations
Review status
Source-backed
Revision
v3 ยท 9,027 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
May 17, 2026
Sources
47 citations
Review status
Source-backed
Revision
v3 ยท 9,027 words
Add missing citations, update stale details, or suggest a clearer explanation.
Runway Gen-4 is the fourth-generation video generation model developed by Runway (company), announced and released on March 31, 2025. The model introduced a reference image conditioning system that allows characters, objects, and environments to remain visually consistent across separate video generations without any fine-tuning or retraining. Runway described Gen-4 as a step toward simulating real-world physics in synthetic video, with improved motion dynamics compared to its predecessor, Gen-3 Alpha.
Gen-4 can produce clips of 5 or 10 seconds at up to 720p resolution (with optional 4K upscaling on paid plans) at 24 frames per second. A faster companion model, Gen-4 Turbo, arrived days later in April 2025, offering approximately five times the generation speed at a lower credit cost. Developer API access for Gen-4 Turbo opened in April 2025, Gen-4 References followed on April 30, 2025, and Gen-4 Image became available via the API on May 16, 2025. An in-context video editing model called Aleph, closely related to the Gen-4 family, launched on July 25, 2025. A successor model, Gen-4.5, arrived December 1, 2025, adding native audio, multi-shot sequencing, and clip lengths up to 60 seconds, alongside Runway's first General World Model.
The Gen-4 release coincided with a $308 million Series D funding round announced on April 3, 2025, which valued Runway at roughly $3 billion. By that point the company had also secured its first Hollywood studio deal, a custom-model partnership with Lionsgate announced September 18, 2024, and was actively pursuing similar arrangements with other studios and networks. By mid-2026 those partnerships also included AMC Networks, Havas Group, Getty Images, and an Adobe Firefly integration.
Gen-4 is positioned as a production-grade generative video tool aimed at filmmakers, advertisers, and other creative professionals. The model accepts an image and a text prompt and returns a short clip in which the image typically functions as the opening frame. The big change from earlier Runway models is the reference image system, which lets a user attach up to three additional images to constrain how a character, object, or environment should look in the output. That detail, prosaic as it sounds, was the gap blocking serious narrative work with previous video models, since a character generated in one shot rarely resembled the same character in the next.
The release also marked Runway's first attempt to be evaluated on the same axes as established VFX work rather than on novelty alone. Reviewers at outlets including TechCrunch, VentureBeat, and SiliconANGLE described the model as "one of the highest-fidelity AI-powered video generators yet" and noted that the consistency capability addressed a problem that had blocked AI video from serious narrative use until then. By early 2026 the model and its successor, Gen-4.5, held the top spot on the public Video Arena leaderboard, ahead of Sora 2 and Veo 3 in blind comparisons.
Runway (company) was founded in 2018 and became a prominent name in AI video generation with a series of incrementally more capable models.
Gen-1 appeared in February 2023 as a video-to-video system. It applied the style or composition of an image or text prompt to the structure of an input video rather than generating video from scratch. The model was practical for style transfer tasks but could not produce novel content without a source clip. In demo reels, Gen-1 turned street footage into claymation and stacks of books into night-time cityscapes.
Gen-2, released in beta on Discord in March 2023 and made available in the browser and on iOS in June 2023, opened up text-to-video and image-to-video generation. Gen-2 attracted significant attention as one of the first accessible commercial text-to-video products, though output length was short and motion often appeared floaty. Time named Gen-2 to its 200 Best Inventions of 2023.
Gen-3 Alpha launched on June 17, 2024 and represented the first model Runway trained on a purpose-built infrastructure for large-scale multimodal training. It could produce 10-second clips from text, images, or video inputs. A speed-optimized variant, Gen-3 Alpha Turbo, followed in August 2024 at a lower credit cost and became the default workhorse for most Runway users through the second half of 2024. But the model still struggled with maintaining consistent character appearance or object identity when the same subject needed to appear across multiple separately generated clips. That cross-clip consistency gap was the central problem Gen-4 was designed to address.
Gen-3 Alpha was also followed by a steady drip of feature releases. Act-One, released on October 25, 2024, generates expressive character performances inside Gen-3 Alpha from a single driving video and a character image, without motion capture or rigging. The lip sync tool, introduced earlier in 2024, eventually supported clips up to 45 seconds long and could sync up to four characters in a single image or video. Frames, Runway's stylistic image base model, arrived on November 25, 2024 and was later folded into the Gen-4 Image lineage.
The Gen-4 family rolled out across roughly four months in 2025, with related research, tooling, and a successor model extending the platform through the rest of the year and into 2026.
| Date | Release | Notes |
|---|---|---|
| March 31, 2025 | Gen-4 (video) | Announced and released on the same day to paid customers via the web app. |
| April 7, 2025 | Gen-4 Turbo | Faster, cheaper variant; 10-second clip in roughly 30 seconds. |
| April 2025 | Gen-4 Turbo on Runway API | Programmatic access for developers. |
| April 30, 2025 | Gen-4 References | Multi-image reference conditioning rolled out to paid subscribers. |
| May 7, 2025 | Gen-4 Image free-tier | Text-to-image and References extended to free users. |
| May 16, 2025 | Gen-4 Image on Runway API | Image generation via API with the same reference system. |
| June 25, 2025 | References update | Improved object consistency and prompt adherence across all tiers. |
| July 25, 2025 | Runway Aleph | In-context video editing model from the Gen-4 research line. |
| August 19, 2025 | Gen-4 Image Turbo | 93 percent of base image quality at 2.5x to 4x lower cost. |
| August 20, 2025 | Act-Two voice | Voice control for character performance generation. |
| December 1, 2025 | Gen-4.5 | Successor model: native audio, multi-shot, up to 60-second clips. |
| December 1, 2025 | General World Model | First in-house world model with Worlds, Avatars, and Robotics variants. |
| December 11, 2025 | Adobe Firefly integration | Runway models embedded inside Firefly and Premiere Pro. |
| June 11, 2026 | AI Festival 2026 (NYC) | Expanded festival covering film, design, fashion, gaming, and advertising. |
Gen-4 itself was announced through the Runway Research blog under the title "Runway Gen-4: AI Video Generation with World Consistency." Coverage on launch day from TechCrunch, SiliconANGLE, and VentureBeat treated the consistency feature as the headline capability and noted that Runway had moved ahead of competitors on character continuity even though it lagged on native audio and text-only generation.
The timing of the release coincided with a period of intense competition in AI video generation. OpenAI's Sora had been temporarily restricting new user signups due to capacity constraints, and Google had not yet publicly released Veo 3. Runway's launch positioned Gen-4 as the most accessible high-fidelity option for professional users at that moment.
The company, which is headquartered in New York, had by that point secured over $230 million in funding from investors including Nvidia and Google. Three days after the Gen-4 release, on April 3, 2025, Runway announced a $308 million Series D funding round led by General Atlantic with participation from Fidelity Management & Research Company, Baillie Gifford, Nvidia, and SoftBank Vision Fund 2. That round valued Runway at approximately $3 billion and brought total funding to about $536.5 million.
Gen-4 generates video clips in 5-second or 10-second durations at 24 frames per second. The default resolution is 1280 by 720 pixels (720p) in a 16:9 aspect ratio. Other aspect ratios are supported, including 9:16 (vertical), 1:1 (square), 4:3, 3:4, and 21:9 (ultrawide). Output files are delivered as MP4 (H.264) or GIF. Upscaling to 4K is available on Standard plans and above.
Unlike pure text-to-video models, Gen-4 is primarily an image-to-video system. Users supply an image as a starting frame and then describe the motion, camera movement, or scene transformation they want. Text prompts support subject motion descriptions, explicit camera movement instructions (pan, zoom, tilt, tracking), style descriptors, and speed or temporal cues such as slow motion. The requirement for an image input distinguishes Gen-4 from systems like Sora 2, which can produce video from text alone, though many users find image conditioning gives more predictable control over the output. Gen-4.5, released in December 2025, added a higher-quality text-to-video mode.
Runway described Gen-4 as representing "a significant milestone in the ability of visual generative models to simulate real-world physics." In practice, the model shows improved rendering of secondary motion effects: hair movement in wind, fabric draping and rippling, water surface dynamics, and environmental particle effects such as dust or smoke. These improvements over Gen-3 Alpha are noticeable in clips involving physical interaction between objects or characters and their environment. Reviewers writing about Gen-4 in mid-2025 noted that the rate of obvious artifacts had dropped from "most generations have issues" to "most generations are clean."
That said, Gen-4 is not a physics simulator in any formal sense. It learns motion patterns from training data rather than from a physical model. Complex interactions, such as two people wrestling or liquid pouring with precise volumetric behavior, can produce artifacts. Fire and water effects in particular sometimes look artificial compared to live action footage. Intricate limb movements during high-speed action also remain a weak point, with occasional deformation in hands, feet, or faces under fast motion. Close-up shots of hands are still unreliable; Gen-4 is better than Gen-3 in this regard, but precise interactions of fingers with objects often require multiple generations to get a usable take. Gen-4.5 closed part of that gap with what Runway calls its "World Engine" representation, an internal state that explicitly tracks gravity, momentum, and object permanence between frames.
Gen-4 responds to camera movement language in text prompts with reasonable fidelity. Users can specify dolly moves, orbiting shots, Dutch angles, and tracking shots. The model supports multiple visual styles, including live-action realism, stop-motion, cel animation, VFX compositing styles, and cinematic film grain. These style parameters can be layered with reference images to produce consistent stylized output across a sequence of clips.
The model includes a Motion Brush tool inherited from Gen-3 Alpha that lets users paint regions of the input image and assign motion intent to those regions. This is essentially per-pixel motion control: a user can specify that a particular character moves left while a background element drifts to the right, with the rest of the frame held mostly static. Reviewers have repeatedly cited Motion Brush as a control feature no competing commercial video model offers at the same granularity.
The reference image system is the capability that most differentiated Gen-4 from prior models at launch. Users can supply up to three reference images alongside a text prompt. The model extracts visual characteristics from those images, including facial identity, clothing details, body proportions, object shapes, surface textures, and environment style, and applies them as constraints on the generated output.
In practice this means a filmmaker can photograph or render a character once and then generate multiple clips of that same character in different locations, under different lighting, or performing different actions, without the character's appearance drifting between generations. Runway's research page described this as enabling "consistent characters across endless lighting conditions, locations and treatments" from a single reference image.
When multiple reference images are used, each image is assigned a label in the prompt (for example, "image_1", "image_2", "image_3") and the user's text describes how each reference should influence the output. One image might supply the character's face while another supplies the background environment or a specific prop. The labels allow the model to resolve which visual elements should come from which source.
The reference system does not require fine-tuning or any model retraining. Conditioning happens at inference time, which makes the workflow practical for production pipelines where subjects change from project to project. Earlier approaches to character consistency in video generation typically required per-subject fine-tuning passes that could take hours and were cost-prohibitive at scale.
Beyond character identity, the reference image conditioning system can transfer artistic styles. Uploading a frame from a film, a painting, or a graphic novel page as a reference, and then describing desired content in the text prompt, causes Gen-4 to produce output that adopts the visual language of the reference: its color palette, lighting approach, texture rendering, and compositional tendencies. This makes Gen-4 useful for producing stylistically coherent series of clips without manually specifying every visual parameter in text.
When Runway rolled out Gen-4 References as a dedicated tool on April 30, 2025, the company emphasized that the same conditioning approach used in video could be applied to still image generation. The June 25, 2025 update improved object consistency and prompt adherence, and was released across all subscription tiers. By that point Gen-4 References was the basis for most of Runway's stylistic control surface, replacing earlier per-feature workflows with one image-conditioned pipeline.
Gen-4 Turbo was released on April 7, 2025, a few days after the standard Gen-4 model. It is a speed-optimized variant designed for rapid iteration and higher-volume production workflows.
The key difference between Gen-4 and Gen-4 Turbo is generation time. Gen-4 Turbo produces a 10-second clip in approximately 30 seconds, which is roughly five times faster than the standard Gen-4 model. This speed advantage makes Turbo practical for creative iteration, where a producer wants to try many different prompt variations quickly before committing to a final shot.
The cost difference is also significant. Gen-4 Turbo consumes 5 credits per second of video (50 credits for a full 10-second clip), compared to 12 credits per second for standard Gen-4. On Runway's Pro plan, which includes 2,250 monthly credits, a user can generate approximately 45 full 10-second clips using Gen-4 Turbo compared to around 18 clips with standard Gen-4.
Quality-wise, Gen-4 Turbo is slightly below the standard model in motion smoothness and fine detail, particularly in complex scenes. For most social media and marketing applications the difference is minimal, but for outputs intended for large screens or professional broadcast contexts the standard model remains the better option.
Gen-4 Turbo became available on the Runway API in April 2025. Runway noted at the time that it offered "the same price, scale and reliability of Gen-3 Alpha Turbo but with the state-of-the-art" quality of the Gen-4 generation.
Aleph is an in-context video editing model released by Runway on July 25, 2025. While Gen-4 and Gen-4 Turbo generate new video from image and text inputs, Aleph works on existing footage and transforms it based on text instructions.
Runway described Aleph as "a state-of-the-art in-context video model" capable of performing "a wide range of edits on an input video." Supported operations include adding objects to a scene, removing objects, transforming the appearance of objects already present, changing lighting and time-of-day, altering the visual style of footage, and generating alternate camera angles from a single shot.
The camera angle generation capability is technically interesting. Aleph can infer a scene's spatial geometry from a single video and then synthesize views that appear to come from a different camera position, such as a reverse angle or an aerial perspective, without any additional footage. This is achieved through depth estimation and spatial reconstruction rather than explicit 3D modeling.
When objects are added or removed, Aleph adjusts shadows, reflections, and lighting to match the scene's existing conditions automatically. The model maintains temporal consistency across the full edited clip rather than processing frames independently, which avoids the flickering artifacts that affected earlier video editing approaches.
Aleph is available to all Runway paid users at 15 credits per second of edited video via the API. The model continued to evolve through 2025 and into 2026, with silent updates to scene relighting, multi-object removal, and prompt adherence on the existing endpoint.
Act-Two is Runway's character performance model, released on July 22, 2025 as a successor to Act-One. It takes a driving performance video plus a character reference image and renders a synthesized performance in which the character mimics the driving actor's expressions, head motion, and body language. A voice control add-on, released on August 20, 2025, accepts a separate audio file and synthesizes matching mouth shapes. Filmmakers on Hundred Film Fund grants in late 2025 routinely paired Act-Two performances with Gen-4 environment plates and Aleph cleanup passes as a single pipeline.
Runway announced Gen-4.5 on December 1, 2025, presenting it as a successor rather than a replacement for Gen-4. The model retains the reference image system and the credit-cost structure of Gen-4 but adds three substantive capabilities: native audio generation, multi-shot sequencing, and clip durations of up to 60 seconds with character consistency preserved across the full minute.
Native audio is the most visible addition. Gen-4.5 produces synchronized dialogue, ambient sound, and sound effects alongside the visual stream rather than requiring a separate sound-design pass, with lip sync generated automatically from the audio track. The feature closed the most frequently cited competitive gap with Veo 3 and Sora 2, both of which had shipped with native audio earlier in 2025.
Multi-shot sequencing lets the model plan a sequence of cuts within a single prompt. A user can describe a wide establishing shot, a medium shot, and a close-up in one prompt and receive a single output that transitions through those compositions while maintaining character and environmental continuity. The feature extends the reference image system from shot-level to scene-level consistency.
Duration was extended from 10 to 60 seconds for character-consistent generations, at higher credit cost; Runway recommended Gen-4.5 for hero shots while keeping Gen-4 Turbo as the iteration default. Independent reviews in early 2026 placed Gen-4.5 at the top of public benchmark leaderboards including Artificial Analysis's Video Arena, where it overtook Veo 3.1 and Sora 2 Pro in blind preference tests. Gen-4.5 also surfaced a "Custom Camera" panel that exposes focal length, f-stop, ISO, and shutter angle as numeric controls rather than natural-language descriptions.
Alongside Gen-4.5, Runway released its first General World Model (GWM) on December 1, 2025. A world model is an AI system that constructs an internal representation of an environment and uses it to predict how the environment will evolve in response to inputs. Where Gen-4 generates a video clip and stops, GWM generates an interactive scene that a user can navigate, manipulate, or query frame by frame.
GWM ships in three variants targeted at different production contexts.
| Variant | Purpose | Typical input | Output |
|---|---|---|---|
| GWM Worlds | Explorable scene generation | Reference images, text prompt | Interactive environment that responds to camera and lighting commands |
| GWM Avatars | Speaking character generation | Character image, audio track | Lip-synced performance with expressive facial motion |
| GWM Robotics | Synthetic training data for robots | Scene description, robot model | Multi-camera video sequences for behavior cloning and RL |
GWM Worlds is positioned as a previs and virtual location-scouting tool: a production designer specifies an environment, generates it once, then moves a virtual camera through it to plan coverage. GWM Avatars is a step beyond Act-Two, since it accepts only an audio file and a single character image. GWM Robotics targets robotics research, where physically accurate synthetic data is in high demand for training manipulation and locomotion policies. The world model is built on the Gen-4.5 backbone, and Runway has framed it as part of a research roadmap that treats video generation as a special case of broader world simulation.
Runway has not published a detailed technical paper for Gen-4. The company's public statements describe the model as built on a transformer-based diffusion backbone with separate visual and temporal encoders, trained on a curated dataset of image and video content. Training data sources, parameter counts, and architectural specifics remain undisclosed, partly for competitive reasons and partly because Runway is one of several generative AI companies facing copyright litigation over training data sources.
The company's research lineage does include published work. Patrick Esser, a principal research scientist at Runway, was a co-author on the 2021 paper "High-Resolution Image Synthesis with Latent Diffusion Models," which introduced the latent diffusion architecture later used by Stable Diffusion. Gen-4 almost certainly relies on a related approach extended to handle temporal frames and reference-image conditioning. With the Gen-4.5 release, Runway disclosed slightly more, describing the model as benefiting from "significant advances in both pre-training data efficiency and post-training techniques" without releasing parameter counts or dataset details.
Runway's developer API gave programmatic access to Gen-4 Turbo in April 2025. The API allows applications to submit image and text inputs and receive generated video as output, enabling integration of Runway video generation into third-party products, platforms, and workflows without going through the Runway web interface.
Gen-4 Image, a related model capable of generating still images using the same reference conditioning system as Gen-4 video, became available on the Runway API on May 16, 2025. A faster variant, Gen-4 Image Turbo, followed on August 19, 2025, delivering 93.3 percent of standard Gen-4 Image quality at 2.5 to 4 times lower cost. Gen-4.5 was added to the API in December 2025, with Aleph, Act-Two, and the GWM Avatars endpoint reachable through the same authentication flow.
Credits purchased through the developer portal cost $0.01 each. Volume discounts apply for larger purchases (for example, 275,000 credits for $1,250). API billing is pay-as-you-go, meaning developers are not required to subscribe to a monthly plan and pay only for the credits they consume.
Runway also released an MCP (Model Context Protocol) server integration in June 2025, allowing the Runway API to be invoked directly from Claude and other AI assistant environments that support MCP. In December 2025 Runway announced an integration with Adobe Firefly that surfaces the Gen-4 family inside Adobe's creative apps, including Premiere Pro and Photoshop. The Firefly integration treats Runway as one of several model providers in Adobe's expanding multi-model lineup, alongside Adobe's own Firefly Video model and Veo 3.
Runway operates on a subscription credit model. As of 2026, the platform offers four consumer-facing tiers plus enterprise custom pricing.
| Plan | Monthly cost (annual billing) | Monthly credits | Gen-4 seconds | Gen-4 Turbo seconds |
|---|---|---|---|---|
| Free | $0 | 125 (one-time) | No Gen-4 access | No Gen-4 access |
| Standard | $12/user | 625 | ~52 seconds | ~125 seconds |
| Pro | $28/user | 2,250 | ~187 seconds | ~450 seconds |
| Unlimited | $76/user | 2,250 + unlimited explore | ~187 seconds (priority) | ~450 seconds (priority) |
| Enterprise | Custom | Custom | Custom | Custom |
The Unlimited plan's "unlimited explore" mode queues generations at a lower priority than paid-credit jobs, which can extend wait times during peak usage. The 2,250 monthly credits on the Unlimited plan can still be used for priority generations at the same rate as the Pro plan.
Credits do not roll over between billing periods. Gen-4 consumes 12 credits per second, Gen-4 Turbo consumes 5 credits per second, Gen-4.5 consumes 18 credits per second, and Aleph consumes 15 credits per second via the API. 4K upscaling is available on Standard plans and above.
Via the developer API, credits cost $0.01 each and can be purchased in any quantity. A 10-second Gen-4 generation costs $1.20 in API credits; the same clip via Gen-4 Turbo costs $0.50. The MindStudio analysis published shortly after the Turbo release calculated that the cost difference made Turbo the natural default for most production work, with the standard model reserved for hero shots or final-quality renders.
Runway's pricing structure has come in for both praise and criticism. Reviewers note that the credit-per-second model is the most predictable in the category for studios managing project budgets. They also note that credits expire at the end of each billing period, which penalizes users with seasonal or sporadic workloads.
In late 2025 Runway repositioned its subscription as a multi-model platform rather than a single-vendor tool. A single paid subscription now provides access to Runway's own models (Gen-4, Gen-4 Turbo, Gen-4.5, Aleph, Act-Two, GWM) along with third-party models including Google's Veo 3.1, Kuaishou's Kling 2.1 and Kling 3.0, ByteDance's Seedance 2.0 and Seedream, and Black Forest Labs' FLUX image models. Credits are shared across the marketplace, with each model assigned a per-second or per-image rate. The move responded to multi-model production workflows in which agencies render the same shot through several models and pick the best output. Trade press in early 2026 referred to the strategy as the "App Store moment" for AI video, with Runway operating as the storefront layer for an expanding model catalog.
For users already familiar with Gen-3 Alpha, the practical differences with Gen-4 are concentrated in character and object consistency, motion realism, and the underlying tool surface.
| Capability | Gen-3 Alpha | Gen-4 |
|---|---|---|
| Max clip length | 10 seconds | 10 seconds |
| Output resolution | 720p | 720p (4K upscale) |
| Text-only input | Yes | Limited |
| Reference images per generation | None | Up to 3 |
| Character consistency across clips | Weak | Strong |
| Motion fidelity | Moderate | Strong |
| Native audio | No | No |
| Credits per second (standard / Turbo) | 10 / 5 | 12 / 5 |
Gen-3 Alpha is still available for users who prefer its specific aesthetic or who have built workflows around it. For most new work in 2025, however, Gen-4 has become the default. The combination of reference-driven consistency and improved motion realism justifies the slightly higher per-second cost for shots that need to integrate with other footage or maintain identity across a sequence.
The primary competitors in high-fidelity AI video generation through mid-2025 were OpenAI's Sora 2, Google's Veo 3, and Kuaishou's Kling 2.1, with MiniMax's Hailuo and Luma's Dream Machine as secondary options. By mid-2026 the field had narrowed: OpenAI announced in March 2026 that the Sora web and app experiences would be discontinued on April 26, 2026, with API access ending September 24, 2026, leaving Veo 3.1, Kling 3.0, and Runway as the leading commercial-grade options.
| Feature | Runway Gen-4 / Gen-4.5 | Sora 2 | Veo 3.1 | Kling 3.0 | Hailuo |
|---|---|---|---|---|---|
| Max clip length | 10s (Gen-4) / 60s (Gen-4.5) | 10-25 seconds | 8 seconds | 10 seconds | 6-10 seconds |
| Output resolution | 720p (4K upscale) | Up to 1080p | Up to 4K | Up to 1080p | Up to 1080p |
| Native audio | Yes (Gen-4.5) | Partial | Yes | Yes | No |
| Reference image conditioning | Yes (up to 3) | Yes (Cameos) | Limited | Limited | Limited |
| Character consistency | Strong | Strong | Moderate | Moderate | Moderate |
| Physics realism | Very good (Gen-4.5) | Very good | Good | Very good | Moderate |
| Generation speed | ~30s (Turbo) | 2-5 minutes | 1-3 minutes | Varies | 1-2 minutes |
| Text-to-video (no image input) | Yes (Gen-4.5) | Yes | Yes | Yes | Yes |
| Starting price | $12/month | $20/month (sunsetting) | $249.99/month (AI Ultra) | Free tier | Free tier |
Runway Gen-4's clearest strengths are character consistency across multiple generations and generation speed via the Turbo variant. Sora 2 showed superior physics simulation for water, cloth, and rigid body dynamics, and was often praised for temporal consistency in longer shots, although OpenAI's decision to wind down the standalone product in 2026 removed it from the active commercial field. Veo 3.1 leads on prompt adherence, native audio quality, and 4K output. Kling 3.0 offers competitive physics and is often considered the strongest image-to-video model for motion realism; Hailuo is praised for fluid character motion but has more variable output quality.
Runway Gen-4's lack of native audio was a notable gap relative to Veo 3 and Kling at launch. The gap was closed with Gen-4.5 in December 2025, which shipped native dialogue, ambient sound, and synchronized music as a single multimodal output. Independent blind aesthetic comparisons in early 2026 placed Gen-4.5 ahead of both Sora 2 Pro and Veo 3.1 on overall quality scores, with Sora 2 still leading on temporal consistency in long shots and Veo 3.1 leading on prompt fidelity and audio quality.
Pricing favors Runway Gen-4 for many production workflows. Veo 3.1 requires a Google AI Ultra subscription at $249.99 per month, well above Runway's Pro tier at $28 per month. Following the Sora 2 sunset, Runway moved to absorb displaced Sora users by offering migration credits and bundled access to Veo 3.1 and Kling 3.0 through its marketplace.
On September 18, 2024, Lionsgate and Runway announced a first-of-its-kind partnership in which Runway would train a new generative video model on Lionsgate's library of more than 20,000 film and television titles. Lionsgate executives framed the deal as a way to augment, not replace, existing creative workflows, and described expected use cases that included pre-visualization, storyboard generation, background and environment creation, and certain categories of visual effects work such as action sequences and explosions.
The Lionsgate deal was the first formal partnership between Runway and a Hollywood studio. It is not a data licensing deal in the conventional sense: Lionsgate's content is not used to train Runway's general consumer video models. The Lionsgate library trains a separate custom model that only Lionsgate and its filmmakers can use. Lionsgate retains ownership of the content that feeds the model and of any outputs generated from it.
Michael Burns, Vice Chair of Lionsgate, told The Hollywood Reporter that the studio expected the AI tools to deliver "transformational impact" on production and saw the partnership as a path to significant cost savings, potentially in the tens of millions of dollars across the slate. Burns added that the studio viewed AI as "a great tool for augmenting, enhancing and supplementing our current operations."
Runway founder and CEO Cristobal Valenzuela framed the deal as a model for how Hollywood and AI companies could work together: "We're committed to giving artists, creators and studios the best and most powerful tools to augment their workflows."
In November 2024, Lionsgate confirmed at an earnings call that filmmakers across its slate were already using the AI tooling, although the studio did not name specific projects. Coverage in Variety and Deadline highlighted that the partnership was structured to keep the company's library separate from the wider training data pool that powers Runway's consumer products, partly as a hedge against the ongoing copyright litigation around generative model training data. By early 2026, trade reports referred to the Lionsgate custom model as "IP-locked" and described it as a template for subsequent studio partnerships.
In June 2025, AMC Networks became the first cable broadcaster to sign a formal partnership with Runway. AMC will use Runway's tools in two main contexts: pre-visualization of upcoming shows and seasons during early development, and marketing asset creation for promotional materials such as posters, thumbnails, and digital campaign images.
Kim Granito, AMC Networks executive vice president, told the Hollywood Reporter that the company saw Runway as a way to experiment with visual directions early in the development cycle without committing to expensive set design or test shoots. AMC has stated that for the time being it does not plan to include AI-generated imagery in final shows; the immediate focus is on internal workflows and promotion. The partnership was extended in early 2026 to broader marketing applications across the AMC portfolio, with access to the Gen-4.5 and GWM Avatars endpoints added.
In December 2025 Runway and Adobe announced an integration surfacing the Gen-4 family inside Adobe Firefly, with text-to-video, image-to-video, and Aleph endpoints reachable from Premiere Pro and Photoshop. Adobe positioned the partnership as part of Firefly's evolution into a multi-model creative surface alongside Firefly Video and Veo 3. Runway has also signed direct deals with several advertising agencies, most notably Havas Group in late 2025. By mid-2026 enterprise customers including Getty Images, Lionsgate, AMC Networks, and Havas accounted for a substantial share of annual recurring revenue.
Madonna's creative team used Runway to generate AI visuals for sections of her Celebration Tour, which wrapped on May 4, 2024 with a free concert on Copacabana Beach in Rio de Janeiro. The most prominent AI-generated visuals appeared during the performance of "La Isla Bonita," where conventional computer graphics had not landed the desired mood.
Sasha Kasiuha, Madonna's creative director, told Harbor Picture Company that the AI tooling was important because the team was on a tight production deadline and wanted to iterate on visual concepts without commissioning new CGI work for each variation. The director generated more than 700 video outputs for "La Isla Bonita" using text-to-video prompts like "surreal sunset, clouds," then selected and edited the best frames into the final stage visual. Following the success of that sequence, the AI workflow was extended to "Take a Bow," which used a more narrative approach inspired by surrealist painters including Salvador Dali and Rene Magritte.
The Madonna engagement was an early high-profile validation of generative video tools in a touring production context. It sat alongside the Lionsgate deal and the AMC Networks partnership as evidence that Runway's tooling had begun to find serious commercial use outside of social media and short-form content.
Runway has built out cultural partnerships in parallel with its commercial deals. In May 2024, the company announced a programming partnership with the 23rd annual Tribeca Festival in New York. The collaboration produced a curated shorts program called Human Powered that screened AI-assisted short films and music videos at the festival.
In August 2025, Runway and IMAX announced a partnership to screen the ten finalist films from the 2025 AI Film Festival exclusively in IMAX cinemas across the United States from August 17 to August 20, 2025. The screenings took place in New York, San Francisco, Los Angeles, Chicago, Seattle, Dallas, Boston, Atlanta, Denver, and Washington DC, with 40 screenings in total. Hollywood Reporter described the deal as the first time AI-assisted short films had been screened across IMAX cinemas at that scale. The IMAX partnership was renewed for the 2026 festival cycle with a planned expansion to additional markets in the US, Canada, and the UK.
In September 2024, Runway announced The Hundred Film Fund, a $5 million grant program (with potential to grow to $10 million) designed to back 100 films that use AI in their production. Grants range from $5,000 to more than $1 million per project, alongside up to $2 million in Runway credits, and cover features, shorts, documentaries, experimental work, and music videos. Applicants must use Runway tooling to some meaningful degree, but Runway does not take ownership of the intellectual property created under the program. Decisions are typically returned within 14 days of submission and the production window is capped at 12 months from the award.
Gen-4 has been used or implicated in a range of high-profile creative outputs since its release, with varying degrees of commercial visibility.
Coca-Cola revived its "Holidays Are Coming" Christmas campaign in 2024 using a fully AI-generated workflow, with three studios (Secret Level, Silverside AI, and Wild Card) producing three distinct variants. The campaign used a mix of generative tools including Runway, Kling, and Luma Labs Dream Machine. Production reportedly involved 10,000 frames and 5,000 video segments. The 2024 ad drew significant backlash online, with critics describing it as "soulless" and a poor homage to the 1995 original. Coca-Cola returned to the AI-driven approach for the 2025 campaign with a similar mixed response. The production team generated tens of thousands of clips and selected a small fraction for the final commercials, which suggests that human curation overhead remained substantial even with the more controllable Gen-4 reference workflow.
The annual Runway AI Film Festival (AIFF) has become the most visible showcase for narrative work produced using Runway tooling. The 2025 edition received approximately 6,000 submissions and was held at Alice Tully Hall in Lincoln Center, New York, on June 5, 2025. The top prize went to "Total Pixel Space," a nine-minute essayistic short directed by Jacob Adler. The film earned $15,000 in cash and 1,000,000 Runway credits. Second place went to "Jailbird" by Andrew Salter, a film told from a chicken's perspective inside a United Kingdom prison; third place went to "One" by Ricardo Villavicencio and Edward Saatchi, a futuristic interplanetary story.
The 2025 jury panel included filmmakers Gaspar Noe, Harmony Korine, and Jane Rosenthal, among others. Variety, Deadline, and The Hollywood Reporter all covered the event. The festival's growth from approximately 300 submissions in 2023 to 3,000 in 2024 to 6,000 in 2025 has been frequently cited as evidence of broader filmmaker engagement with AI tools.
Beyond the festival, a growing list of brand commercials in 2024 and 2025 used Runway tooling in some part of their production pipelines. The specifics of which model variant was used (Gen-3 Alpha, Gen-3 Alpha Turbo, Gen-4, Gen-4 Turbo, Gen-4.5) often were not disclosed by the brands. Agencies typically cite Runway alongside multiple other generative video models in their production credits, reflecting a multi-model workflow rather than a single-tool dependence.
Runway Studios is the production arm of Runway, established in 2023 to develop and produce films, episodic content, and other media using the company's generative tooling. The studio's stated purpose is both creative (producing original work) and commercial (showing what Runway's tools can do in practical production contexts).
The AI Film Festival is the most visible Runway Studios initiative. The first festival in 2023 received roughly 300 submissions. The second in 2024 received approximately 3,000 submissions, including the "Human Powered" Tribeca shorts program. The third in 2025 received 6,000 submissions and was held on June 5, 2025 at Alice Tully Hall at Lincoln Center, followed by IMAX cinema screenings from August 17 to August 20, 2025.
The 2026 edition, announced in January 2026 and rebranded as the AI Festival, expanded scope beyond film to include design, new media, fashion, gaming, and advertising. The New York event is scheduled for June 11, 2026 at Alice Tully Hall, with a companion Los Angeles event on June 18, 2026. Submission volume in 2026 was projected to exceed 10,000.
Gen:48, a recurring 48-hour challenge run by Runway Studios, gives filmmakers a fixed prompt or theme and 48 hours to produce a short film using Runway tools. The competition serves as a recruitment and engagement pipeline for the larger festival and the Hundred Film Fund. Runway has said it intends to expand Runway Studios into more conventional film and animation production, with the April 2025 Series D round earmarked in part to grow the studio's capacity. Runway Studios productions screened at the 2026 Tribeca Festival included two feature-length pilots and a six-episode animated series, all made with a combination of Gen-4.5, Aleph, and Act-Two.
Coverage at launch was largely positive about the character consistency capability, which reviewers recognized as a genuine step beyond Gen-3 Alpha. TechCrunch described the model as "impressive" in its March 31, 2025 report on the release. Reviewers also noted that Gen-4 set a new standard for cinematic output quality among accessible consumer-facing tools, with one analysis calling it the "benchmark for cinematic AI video quality in 2025."
Critical perspectives focused on several recurring issues. The 10-second clip limit was a genuine constraint for narrative filmmaking, which requires longer continuous shots. Stitching many 10-second clips together introduced consistency and pacing challenges that required additional editing work. The absence of native audio was noted as a competitive disadvantage relative to Veo 3 and Kling. The credit cost for standard Gen-4 (12 credits per second) was described as expensive for exploratory or high-volume use, which is why many users shifted to Gen-4 Turbo. The arrival of Gen-4.5 in December 2025 addressed most of these criticisms simultaneously by extending durations, adding native audio, and improving physics fidelity, leaving the credit-cost question as the main remaining critique.
Bloomberg reporting around the April 2025 Series D round emphasized that Runway's strategy of cultivating studio partnerships (Lionsgate, AMC Networks, IMAX) was part of a deliberate effort to position the company as a Hollywood-aligned vendor rather than a Hollywood disruptor.
The broader context of AI-generated video attracted criticism that applied to Runway alongside other companies in the space. Runway has declined to disclose the sources of its training data, citing competitive concerns. In July 2024, 404 Media reported that an internal Runway spreadsheet leaked by a former employee appeared to show that the company had scraped training data from thousands of YouTube channels including those of The New Yorker, Vice News, Pixar, Disney, Netflix, Sony, and individual creators such as Casey Neistat, Sam Kolder, and Marques Brownlee. According to the report, employees scraped the videos with youtube-dl through proxy servers to evade YouTube's anti-scraping measures. Runway co-founder Anastasis Germanidis told TechCrunch in a separate interview that the company used "curated, internal datasets" without commenting on the specific allegations. Runway is also a defendant in Andersen v. Stability AI, the multi-plaintiff copyright case filed in January 2023 by visual artists against several AI image generation companies.
Several limitations of Gen-4 were clear at launch and remained through its production lifetime, although the Gen-4.5 update in December 2025 narrowed several of them.
Clip length topped out at 10 seconds for the original Gen-4 model. Longer sequences required chaining multiple clips together, which demands additional editing work and can produce consistency issues at clip boundaries despite the reference image system. Filmmakers working on narrative shorts in 2025 routinely described the 10-second cap as the single biggest practical constraint. Gen-4.5 extended that limit to 60 seconds for character-consistent generations, which reduced but did not eliminate the issue for longer-form work.
Gen-4 did not support text-only video generation; an input image was required to anchor the generation. Audio was absent from native Gen-4 output, so all sound design had to be added in post-production. Gen-4.5 added both native text-to-video and native audio, bringing Runway to parity on those two capabilities.
Physics simulation, while improved over Gen-3, fell behind Sora 2 and Kling in direct comparisons involving fire, fluid dynamics, and complex rigid body interactions. Fast-motion sequences involving hands, fingers, or feet could produce deformation artifacts. Hand artifacts remain an industry-wide issue that affects every major commercial video model.
The credit system means costs scale with usage in ways that can be hard to forecast. A production that requires many exploratory iterations to land on the right output can consume credits quickly, and monthly credits expire if unused. Repeatability is also limited: identical prompts and reference images do not guarantee identical outputs, which is inherent to diffusion-based generation but can frustrate users who need to reproduce a specific result. The 720p base output (with 4K upscaling as a separate step) was also a constraint relative to Veo 3 and Sora 2 throughout the Gen-4 production lifetime.
Gen-4 attracted adoption across several professional production contexts in 2025.
Pre-visualization (previz) is one of the most common applications. Directors and production designers use Gen-4 to mock up shot ideas before committing to physical setups or expensive CGI. A reference image of a character combined with a text description of a camera move can produce a rough previz clip in under a minute, fast enough to show to a client or investor during a pitch meeting. The Lionsgate and AMC partnerships are explicitly structured around this use case. With the December 2025 arrival of GWM Worlds, the previz workflow extended to interactive scene exploration.
Advertising and branded content production benefit from the character consistency feature. A brand can photograph talent or a product once, then generate multiple stylistically distinct clips of the same subject in different settings without scheduling additional shoots. Independent filmmaking has been a significant adoption driver too, with low-budget productions using Gen-4 to generate VFX shots, environment extensions, and concept shots that would otherwise require expensive CGI vendors. The Hundred Film Fund explicitly targets this segment, and the films that win Runway's AI Film Festival typically come from independent filmmakers.
Social media content creation, particularly for TikTok and Instagram Reels, benefits from Gen-4 Turbo's speed. Educational creators use Gen-4 References to keep a recurring character or environment consistent across a series of short clips, which would have been impractical with earlier models.
Runway was founded in 2018 in New York City by Cristobal Valenzuela, Anastasis Germanidis, and Alejandro Matamala, who met as graduate students at New York University's Tisch School of the Arts Interactive Telecommunications Program (ITP). ITP is a program at the intersection of art and technology, often described as an art school for engineers and an engineering school for artists. Valenzuela earned his media arts degree from ITP in 2018 and remained in residence as a researcher before formally launching Runway with Matamala and Germanidis that same year.
The early Runway product was a browser-based creative tool that let artists and filmmakers experiment with pre-existing machine learning models for image manipulation and effects work. By 2020 the company offered a hosted platform with tools for rotoscoping, green-screen replacement, and image inpainting.
Runway's research line attracted attention beyond product circles in 2021 when researchers from the company co-authored "High-Resolution Image Synthesis with Latent Diffusion Models" alongside collaborators at LMU Munich and Heidelberg University. That paper introduced the latent diffusion model architecture that became the basis for Stable Diffusion. In August 2022, Runway, the CompVis lab, and Stability AI jointly released the first Stable Diffusion checkpoint. Runway later hosted the Stable Diffusion 1.5 model on Hugging Face, one of the most-used open-weight image generation models of 2022 and 2023.
Valenzuela serves as CEO. Germanidis is Chief Technology Officer. Matamala leads design. Patrick Esser, a principal research scientist and co-author on the original latent diffusion paper, leads the applied research line for video generation.
| Date | Round | Amount | Lead | Valuation |
|---|---|---|---|---|
| 2018 | Seed | $2 million | Various | Not disclosed |
| December 2020 | Series A | $8.5 million | Amplify Partners | Not disclosed |
| December 2021 | Series B | $35 million | Felicis Ventures | Not disclosed |
| December 2022 | Series C | $50 million | Felicis Ventures | $500 million |
| June 2023 | Series C extension | $141 million | Google, Nvidia, Salesforce | $1.5 billion |
| April 2025 | Series D | $308 million | General Atlantic | $3 billion |
The June 2023 Series C extension was reported widely as a marker of generative AI's commercial momentum that year, with Google, Nvidia, and Salesforce Ventures all participating. The April 3, 2025 Series D round, led by General Atlantic with Fidelity Management & Research, Baillie Gifford, Nvidia, and SoftBank Vision Fund 2, valued the company at approximately $3 billion and brought total raised capital to roughly $536.5 million. Runway said the funds would support AI research, hiring, and the growth of Runway Studios.
Time magazine named Gen-2 to its 200 Best Inventions of 2023, Gen-3 Alpha to its 200 Best Inventions of 2024, and also named Runway one of its 100 Most Influential Companies in June 2023.
Gen-4's release and Runway's expanding studio partnerships landed during an extended industry debate about AI's place in film and television production. The 2023 Writers Guild of America and SAG-AFTRA strikes both centered, in part, on protections against AI-generated replicas of writers and performers. The agreements that ended those strikes established disclosure requirements for digital replicas of named performers and bargaining requirements for the use of fully synthetic performers.
Runway's positioning has consistently been that the company builds tools for augmenting human creative work rather than replacing it. The Hundred Film Fund, the AI Film Festival, the Tribeca and IMAX partnerships, and the framing of the Lionsgate and AMC deals all reflect that positioning. Critics in the trade press have argued that this framing oversells the augmentation aspect and underplays the displacement risk to roles such as concept artists, storyboard artists, and junior VFX technicians. SAG-AFTRA's 2026 contract renewal negotiations, which began in March 2026, are expected to revisit the disclosure framework, as the Gen-4.5 and GWM Avatars features have substantially improved the realism of synthetic performers since the 2023 contract was signed.
The Brutalist, a 2024 film directed by Brady Corbet, was sometimes erroneously associated with Runway in early coverage. The film's AI use was limited to dialogue post-production work that adjusted Hungarian-language pronunciation by performers including Adrien Brody and Felicity Jones, and was performed using audio tools from a different company (Respeecher). Runway was not involved in that production.