Runway: video generator — what it does, formats and rights to the result

Runway is a cloud service for generating images, video, and audio from text. The developer positions it as a tool for creating content with artificial intelligence. The service's code is closed, and it does not run locally.

What it does

The input can be text, an image, video, or audio. The authors note that there are three ways to pass an image, video, or audio file, and all of them are described in the documentation.

For video generation, the gen4.5 model is available. It accepts an aspect ratio of 1280:720 or 720:1280, as well as any integer duration from 2 to 10. The aspect ratio and duration parameters are required. The prompt for this model can contain up to 1,000 characters.

Among other available models, the developer mentions gemini_omni_flash, seedance2, and aleph2.

For video enhancement, the magnific_video_upscaler_creative model is used. Input videos can be no longer than 30 seconds. The frame rate is 30 fps.

How to use it

The service is available via API. The developer provides official SDKs for Node (the @runwayml/sdk package) and Python (the runwayml package). The authors recommend using the SDKs rather than direct HTTP requests, since the SDKs handle version headers, retries, and polling of task status. There is also a React SDK for working with characters.

The service does not release its own model or weights, and local running is not possible.

Rights to the result

The fact sheet does not specify the terms for transferring rights. This point should be clarified in the contract or the terms of use of the service itself.

Who it suits

The service is aimed at developers who are ready to integrate video generation into their applications via API and SDK. It suits those who need generation from text, image, video, or audio without having to deploy their own infrastructure.

It does not suit those looking for an open-source solution with the ability to run locally, fine-tune models, or work without an internet connection. The 30-second limit for input video during enhancement is also important to consider when planning tasks.

Runway provides access to several generative models through a single API, but the details of working with each of them, including specific limitations and formats, should be clarified in the service's documentation.