Blend two videos together API: what Sume can and cannot layer

Sume cannot blend two videos into one frame: video filter works on one clip, and compose puts one still over one video. Two videos join in sequence on Timeline.

4 min readSume
All posts

Sume has no endpoint that blends two videos into one frame. The video filter takes one clip, and timeline compose layers one still with one video. To combine two videos, place them one after the other as video[] slots on Timeline 1.0, with a transition such as dissolve between them.

Runway's Aug 18, 2026 changelog lists new compositing workflow nodes that blend images and video. This post covers what Sume's docs say about the same job, read 2026-10-01: Timeline compose, Video filter and Timeline 1.0.

Why does the video filter list blend and overlay?

The allowlist in the filter compiler includes split, overlay, hstack, vstack and blend, but the source comment marks them as internal labels only: inputs are still one clip. You can split one clip into two branches and recombine them, for example to build an effect. You cannot feed a second video in; the program has no inputs of its own, and the server wraps the single clip as the source.

What can compose do instead?

Compose takes one still and one video and returns one MP4 with both on screen. operation: "stack" tiles two regions, and operation: "overlay" puts the still on the video with position, width_ratio (0.05-1) and margin_ratio. Fit options are cover, contain, stretch; the docs say blur is not a compose fit.

Which Sume surface fits which combination, from the Sume docs, read 2026-10-01.
GoalSurfaceNote
Effects on one clipPOST /v1/video-filterOne video_url; allowlisted filtergraph
Still over or beside a videoPOST /v1/timeline-1.0/composeoperation is stack or overlay
Video then videoPOST /v1/timeline-1.0/renderOrdered video[], transitions on slots after the first
Two videos in one frameNot offeredCompose requires a still plus a video

How do I get a dissolve between two videos?

Import both clips with POST /v1/media-imports, then send a Timeline render with two video[] slots and an audio spine (audio.duration_seconds plus audio.url, or audio.mode set to "silence"); the second slot carries transition: { "type": "dissolve", "duration": 0.5 }. Transition duration is capped at 1 second and at 50% of the shorter neighbouring slot. Sizing options are covered in Timeline fit modes.

What if I need a picture-in-picture of two videos?

The docs do not describe one. Compose's image.url must probe as a still, otherwise it returns compose_image_not_still. If you need that layout, produce it outside Sume and import the result. For graphics on a presenter clip, see talking-head video with graphics.

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume