Runway ML
Description
Runway is a creative AI platform that lets teams generate, edit, and transform video using a suite of machine-learning models built for production workflows. The following Runway nodes are included: Runway Act-Two, Runway HDR, Runway Image to Video, Runway Text to Image, Runway Text to Video, Runway Upscale Image, Runway Upscale Video and Runway Video to Video.
Note: Internet access, a Runway Dev account and agreement to Runway ML’s Terms of Use are required to use Runway ML. Sign in, go to Dashboard > API Keys, copy the key, press Generate in Silhouette, and choose Enter API Key to add it.
Node Group
AI/ML, Runway ML.
Workflow
Apply a Runway node, enter your text prompt, and click Generate. The Runway task appears in the Tasks window and runs in the background. When it finishes, click Show Results to open the Asset Browser, where you can view or play the generated media.
The output is automatically saved in the project’s assets/runway folder.
In the Asset Browser, the icons in the lower-left corner let you create a source, create a source with a source node, or delete the asset. The … Options menu gives you access to the file location or lets you export the media.
Controls
Runway Act-Two
Animates your character using the performance from the main input clip. The second input must contain a clear, unobstructed face so the system can accurately track and transfer expressions. Act-Two applies the acting from the source clip to your character’s face and, when available, body, creating a new performance driven entirely by the original footage.
Runway HDR
Runway HDR converts standard dynamic range (SDR) video to high dynamic range (HDR) using Runway’s Ruby model. It extends brightness and color into the HDR range while retaining the input video’s resolution. No text prompt is required. Runway accepts SDR clips up to 30 seconds long and up to 4K resolution.
Runway Image To Video
Generates a short video from a single still image. You provide one image, and Runway animates it into a moving clip by inferring motion, depth, and scene dynamics. The result is a video that preserves the look of the original image while adding natural camera movement or environmental motion.
Runway Text To Image
Creates an image from a written description. You enter a prompt describing what you want to see, and Runway generates a still image that matches your text. It is designed for concept art, visual exploration, and quick ideation, giving you a fast way to turn written ideas into visuals without needing reference images or additional inputs.
Runway Text To Video
Creates a video from a written description. You provide a text prompt, and Runway generates a short clip that visualizes your idea with motion, lighting, and scene dynamics inferred from the prompt. It is designed for rapid concept development, story exploration, and previsualization, giving you a fast way to turn written ideas into moving imagery without needing reference footage or additional inputs.
Runway Upscale Image
Upscale Image increases the resolution of a single image using AI super-resolution. It enlarges the image, sharpens edges, reconstructs detail, and reduces artifacts while preserving the original composition. The node uses the Magnific Precision Upscaler model, which is designed for high-quality image enhancement.
Runway Upscale Video
Upscale Video increases the resolution of an existing video and produces a higher-quality version that stays faithful to the source. It enlarges each frame, sharpens edges, restores detail, and reduces noise while preserving motion, timing, and aspect ratio. The node uses Runway’s Magnific Video Upscaler model.
Runway Video to Video
Transforms an input video into a new stylized or reimagined version while keeping the original motion and timing. You provide a source clip and a text prompt describing the look or style you want, and Runway generates a new video that follows the same movement but with updated visuals. It is designed for concept development, look exploration, and rapid iteration, giving you a fast way to restyle existing footage without manual animation or rotoscoping.