Runway
Runway is a cloud platform for generating and editing video, images and audio with AI. It combines in-house models such as Gen-4.5 and Aleph 2.0 with third-party models, a developer API and world-model tooling for robotics.
What is Runway?
Runway is the product of Runway AI, Inc., a Delaware company founded in 2018 out of NYU's Tisch School of the Arts by Cristobal Valenzuela, Alejandro Matamala and Anastasis Germanidis. It began as a creative toolkit and now describes itself as building Real-World Intelligence: general world models meant to understand, simulate and act in the world. The company reports offices in New York, San Francisco, Seattle, London, Tel Aviv and Tokyo.
Three products sit on the same model stack. Runway Creative is the cloud workspace where most people meet the tool, generating and editing video, images and audio in one place. It runs in-house models such as Gen-4.5, Aleph 2.0, Act-Two and Gen-4 Image alongside third-party models surfaced inside the product, among them Seedance, Kling, Nano Banana, Seedream, Gemini Omni, FLUX, Claude Opus and Veo. Runway Dev exposes the same catalogue through an API, with model routers, workflows that chain models and modalities into custom endpoints, and Characters for real-time video agents carrying custom avatars, custom voices, a knowledge base and tool calling. Runway Robotics applies the GWM-1 world model to policy inference, offline policy evaluation, data augmentation and video model licensing, fitted to a customer's own hardware.
Access is broad: a browser application, iOS and Android apps, the API, and an official MCP server that plugs the platform into Claude, ChatGPT, Cursor or Replit. Everything runs on credits, spent per generation according to model, duration and resolution.
The scale claim is 60 million creators, and the partner list is verifiable: Lionsgate, which has also taken an equity interest, NVIDIA, Amazon's House of David, CBS, Publicis, New Balance and Salomon. Runway publishes research, including the latent diffusion paper behind Stable Diffusion, and claims a role in that model's open-source release. Roughly 850 million dollars has been raised across six rounds since 2020, most recently a 315 million dollar Series E in February 2026 led by General Atlantic, and a London European headquarters was announced with a planned 100 million dollar UK investment.
What it does
- Generate video from text, an image, another video or an audio clip
- Edit a single frame and propagate the change across an entire shot
- Create and retouch images: relight, restyle, change backdrop, weather or time of day
- Upscale footage to 4K and remove unwanted elements from a video
- Produce voice-overs, lip sync and custom voices, plus music and generated audio
- Chain several models and modalities into workflows exposed as custom API endpoints
- Run robot policy inference and generate synthetic data inside photorealistic simulation
When to use Runway / When not to
A quick filter to help you decide if Runway is the right fit.
When to use Runway
- Video editors, colourists and post-production teams who need to generate or repair shots without organising a reshoot
- Creative agencies and brand teams producing campaign variations, product films and social content at volume
- Independent creators and freelancers who want frontier video models for around 12 USD a month
- Developers building media pipelines who want several image, video and audio models behind a single API
- Enterprises needing single sign-on, configurable teamspaces and a contractual commitment not to train on their data
When not to use Runway
- Anyone under thirteen, who is barred by both the terms of use and the privacy policy
- Individual users who need their prompts and outputs kept out of model training, since no self-service opt-out is documented
- Organisations required to keep personal data inside the European Union, as processing takes place in the United States
- Teams that need a non-English interface, since English is the only language the site declares
- Users expecting phone or direct email support, since the contact form routes product questions to the Help Center
How to use Runway
A typical end-to-end flow, from setup to results.
- Create an account on the Runway web application, or sign in with a third-party account such as Google
- Start on the permanent free plan, which credits 125 one-time credits that never expire
- Pick a tool, whether video, image, audio or editing, and choose the model you want to run
- Describe the result in a prompt, or upload an image, a video or an audio clip to work from
- Watch the credit cost as you go: 60 credits buys five seconds of Gen-4.5, eight credits one 1080p Gen-4 image
- Upgrade from the Plans and Billing page when you need more, or buy top-up credits in blocks of 1,000
- For code, generate an API key in the developer portal and follow the setup, models and go-live guides
- Assemble workflows that chain several models, then trigger them through your own custom endpoints
- To drive Runway from an assistant, add the Runway MCP connector in Claude, ChatGPT, Cursor or Replit and sign in
- Send product questions to the Help Center, since the contact form explicitly does not handle support
Pros & Cons
Pros
- One workspace and one API for Runway's own frontier models and a wide catalogue of third-party ones
- Security posture documented rather than asserted: a dated trust page, external penetration tests and a private bug bounty
- A named subprocessor list with processing countries, plus a customer Data Processing Addendum
- Outputs stay yours and commercial use is not restricted, a point repeated on the enterprise page
- A genuinely permanent free plan with 125 credits that never expire, and a 12 USD entry tier
- Public and complete API documentation with code examples, plus an official MCP server
- 25% student and educator discount verified through SheerID, with classroom licensing from three seats
Cons
- On individual plans the terms allow training on your prompts and outputs, and no opt-out is documented
- No Article 27 EU representative, and no mention of standard contractual clauses or the Data Privacy Framework
- Processing happens in the United States, with subprocessors in Singapore, Canada, the United Kingdom and Spain
- Retention is described in general terms; only face scans and voiceprints carry a figure, three years
- The only address published on the site is a Delaware registered-agent address in Dover, not an operating office
- Credits do not roll over on Standard and Pro, and the 12 USD headline price requires an annual commitment
- No direct support email on the contact page, and English is the only interface language declared
Pricing & Plans
Runway offers a permanent free plan rather than a time-limited trial: it grants 125 one-time credits that do not expire, 5 GB of asset storage and a selection of generative models. The cheapest paid entry point is the Standard plan at 12.00 USD per month when billed annually, or 15.00 USD per month when billed monthly. Annual commitment carries a 20% discount across every paid tier, and additional credits may be purchased separately in blocks of at least 1,000.
- Free — 0 USD per month — 125 one-time credits
- 5 GB asset storage
- a selection of generative models
- Standard — 12 USD per month billed annually
- 15 USD billed monthly — 625 credits a month
- all image and video models
- 4K upscaling
- no watermarks
- Pro — 28 USD per month billed annually
- 35 USD billed monthly — 2
- 250 credits a month
- custom voices for lip sync and text to speech
- 500 GB storage
- unlimited Topaz AI upscaling
- Max — 76 USD per month billed annually
- 95 USD billed monthly — 9
- 500 credits a month
- one month of credit roll-over
- first access to new models
- ProRes and image sequence output
- HDR up to 16 bit
- Enterprise — custom pricing — bespoke credit packages
- single sign-on
- workspace analytics
- configurable teamspaces
- organisation-wide onboarding and priority support
- Runway Dev — a separate API pricing tab exists on the pricing page
- its amounts are not rendered in static HTML and were not recorded
Data, GDPR & hosting
A consolidated view of how Runway handles your data.
GDPR overview
Concrete GDPR measures exist, though the site never claims outright compliance. The privacy policy, effective 11 May 2026, names Runway AI, Inc. as data controller, sets out its legal bases, and describes EEA and UK rights including complaint to a supervisory authority; requests go to privacy@runwayml.com. The security page states GDPR Readiness and describes protections for European users. A Runway Customer Data Processing Addendum is referenced, and a named subprocessor list with processing countries is published together with a change-notification subscription. Two gaps stand out: no Article 27 EU representative is designated, and neither standard contractual clauses nor a Data Privacy Framework certification is mentioned, even though the policy has users consent to transfer to the United States.
Who owns the data?
Runway's terms of use state that the company claims no ownership of your content, nor of the inputs you submit or the outputs its models return, and that it does not restrict commercial use of those outputs. The enterprise page repeats the point as full ownership of your outputs. In exchange you must warrant that you hold every right needed for what you upload. Three carve-outs matter: stock avatars supplied by Runway remain Runway's, any feedback you send grants the company a perpetual royalty-free sublicensable licence, and your account and API keys are company property in which you hold no interest.
Reuse rights
You may reuse what you generate without asking permission. The terms confirm that Runway claims no ownership of inputs or outputs and places no restriction on their commercial use, so generated video, images and audio can go straight into client work, advertising or a product. The constraint runs the other way. By using the service you grant Runway a non-exclusive, irrevocable, perpetual, worldwide, royalty-free, transferable and sublicensable licence over those same inputs and outputs, explicitly including training and improving its AI models, alongside labelling, classification and content moderation. Enterprise customers are told the opposite in five words, We don't train on your data, but no equivalent exclusion is documented for free or individual paid plans. Named subprocessors including Google, Anthropic, OpenAI, Hive and Databricks may process content for generation or moderation.
Data retention & training
Hosting summary
Runway states in its privacy policy that the company is headquartered in the United States and that using the service means consenting to the collection, processing, maintenance and transfer of information in and to the United States. The security page describes infrastructure spread across multiple physically separated data centres with major cloud providers, protected against DDoS attacks and monitored around the clock, with a public status page. Data travels over TLS 1.2 or above and rests under AES-256, with keys held in managed systems and rotated on schedule; automated backups are tested, and a written disaster recovery and business continuity plan exists. The published subprocessor list names the infrastructure providers, all in the United States: Amazon Web Services, Google Cloud, Modal Labs, Oracle America and CoreWeave. Other subprocessors process content elsewhere, including Runware and VEED in the United Kingdom, Freepik in Spain, Byteplus, Kling AI and Nanonoble in Singapore, and Ideogram in Canada. No dedicated European hosting region is offered.
Where Runway works
Country-level availability.
Not available in
Things to keep in mind
Risks and trade-offs to weigh before adopting Runway.
- Clause 4.4 grants Runway a perpetual, irrevocable, sublicensable licence over your prompts and generated results, model training included: read it before uploading a client's unreleased material
- The promise not to train on customer data appears only on the enterprise page, so free and individual paid users have no documented equivalent
- The terms impose binding arbitration and waive class actions unless you post a written opt-out to Dover, Delaware within thirty days
- Credits create a quiet spending pressure: they reset on Standard and Pro, top-ups start at 1,000, and cancelling mid-term brings no pro-rata refund
- SOC 2 Type II is the vendor's own statement with the report gated behind a trust portal, while ISO 27001 and NIST are cited as inspirations rather than certifications
- Superlatives such as the world's best video model come from Runway alone, with no independent evaluator named anywhere on the site
- Effortless generation can quietly erode the habit of shooting, framing and editing yourself, so treat the tool as an accelerator of craft rather than a replacement for it
Setup & Integrations
Technical difficulty
Low for the product itself: nothing to install, everything runs in a browser or the mobile apps, and an account takes a couple of clicks or a Google sign-in. The enterprise page claims no complex interfaces or technical setup are needed, and the real learning curve is prompting and budgeting credits. The developer path is another matter, since an API key, the setup and go-live guides, model routers and assembled workflows all assume integration skills. The MCP connector takes three steps. Robotics is a project rather than a signup: Runway fits it to your hardware.
Deployment
Apps stores
Integrations
Supported languages
Behind Runway
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Is there a free plan, or only a trial?
What does the cheapest paid plan cost?
How do credits work?
Can I use the videos and images commercially?
Will my content be used to train Runway's models?
Is there an API, and are there mobile apps?
What is the minimum age?
Where is my data processed?
Which certifications does Runway hold?
Can I connect Runway to an AI assistant?
Should you pick Runway?
Runway is one of the few generative video companies with the funding, the research record and the product depth to be treated as infrastructure rather than a novelty. Roughly 850 million dollars raised across six rounds since 2020, a February 2026 Series E of 315 million led by General Atlantic, an equity partnership with Lionsgate and a published research trail give it a durability most tools in this category lack. The product reflects that: one workspace running Runway's own Gen-4.5 and Aleph 2.0 next to a rotating catalogue of third-party models, the same catalogue behind an API with model routers and chained workflows, and a robotics line built on the GWM-1 world model.
The security and privacy documentation is unusually concrete for the category, with a dated trust page, SOC 2 Type II claimed and the report available on request, a named subprocessor list with processing countries, and a customer Data Processing Addendum. Outputs stay yours, and commercial use is not restricted.
The reservation is specific and worth weighing. Clause 4.4 of the terms grants Runway a perpetual, irrevocable licence over your inputs and outputs, explicitly including model training, and no opt-out is documented for free or individual paid plans. The enterprise page states the opposite in five words, which makes the exclusion a matter of contract rather than a setting. Add processing in the United States, no Article 27 EU representative, retention described mostly without figures, and a Delaware registered-agent address as the only one the site publishes.
For creators and agencies who care about capability and own their client relationships, the 12 dollar entry price makes the trade an easy one. For anyone handling confidential or regulated material on an individual plan, read clause 4.4 before uploading.
- Choosing a selection results in a full page refresh.
- Opens in a new window.