Runway / gen 4.5

Overview

Gen-4.5 is Runway’s video generation model. It creates short clips from a written scene description, or it animates a provided first-frame image. It outputs a finished MP4 video that you can edit, stitch, or composite into larger timelines. It’s useful when you need believable motion, clear camera direction, and strong prompt follow-through.

What you can build

• Storyboard clips for pitches, animatics, and previs
• Product and lifestyle b-roll concepts for ads
• Cinematic camera moves like pans, trucking, and handheld drift
• Image animation from a single still frame for social posts
• Mood and style explorations for film, music videos, and games
• Short VFX plates to combine with live action or 3D renders

Inputs

• A written description of the shot, including subjects, action, timing, and camera movement (1 to 1,000 characters).
• Optional first-frame image to animate (JPEG, PNG, or WebP). This switches the workflow to image-to-video.
• A target aspect ratio for the output clip. Common options include 16:9 (1280×720), 9:16 (720×1280), 1:1 (960×960), 4:3 (1104×832), 3:4 (832×1104), and 21:9 (1584×672).
• A clip length between 2 and 10 seconds.
• Optional reproducibility control. Provide a whole-number seed if you want repeatable outputs.
• Optional public-figure moderation sensitivity setting. This can be used to set a lower threshold for public-figure detection.

Outputs

The model returns a generated video clip as an MP4 file. The clip duration matches the requested length (2 to 10 seconds). Output resolution is 720p, using the selected aspect ratio. Runway supports 24 fps and 25 fps outputs.

Recommended settings

• Use 5 seconds for single, simple actions.
• Use 10 seconds when your prompt includes multiple actions with clear sequencing.
• If you provide a first-frame image, pick an output aspect ratio that matches it to reduce center-cropping.
• Set a fixed seed when you need the same clip structure across repeated runs.

Limitations

• Clips are limited to 2 to 10 seconds.
• Output quality is capped at 720p.
• Text-only generations may have fewer supported output sizes than image-conditioned runs.
• If your first-frame image does not match the requested aspect ratio, the system may center-crop it.
• Very small images, blurry images, or heavy compression reduce detail stability and motion quality.
• Inputs with logos, watermarks, or overlaid text can cause failures or degraded results.
• Like many video models, it can show causal errors (effects before causes), object permanence issues (items appear or vanish), and “success bias” where actions succeed unrealistically.

Safety & compliance

Runway moderates both inputs and outputs. Content that violates Runway’s Usage Policy may be blocked.

Key prohibited areas include child sexual abuse material, sexually explicit content and adult nudity, graphic violence and gore, terrorism or violent extremism, hateful conduct, harassment, and content promoting self-harm.

Runway also restricts misuse that violates rights or safety. This includes non-consensual intimate imagery, using another person’s likeness without permission, impersonation for fraud, and some requests that mimic living artists’ styles.

Repeated moderated requests can lead to enforcement actions such as account suspension. Runway also supports provenance signals (including C2PA) to help identify AI-generated content.