Video Generation node

The Video Generation node produces a video clip from a text prompt using a dedicated video model (via Replicate).

Handles

Handle	Purpose
prompt	The text prompt describing the clip
first	The starting frame (image)
last	The ending frame (image)
references	Reference images (image-to-video)

Configuration

• Model — the video model (e.g. minimax/video-01).
• Aspect ratio — e.g. 16:9, 9:16 (where supported).
• Resolution — e.g. 1280x720 (where supported).
• Duration — clip length in seconds (where supported).
• FPS — frames per second (where supported).
• Seed — deterministic output (where supported).

Important: model capabilities

Video models differ in what inputs they accept:

• MiniMax Hailuo (minimax/video-01) — text + first frame + reference image.
• Seedance 1 Pro (bytedance/seedance-1-pro) — text/image-to-video, start and end frame, fps + seed.
• Luma Ray 2 (luma/ray-2-720p) — text + start/end image (no references).
• Wan 2.1 (wan-video/wan-2.1-1.3b) — text only.

The node warns — and the engine fails fast with an actionable message — when you wire references or frames into a model that cannot consume them.

Example — text + first frame

prompt: "A drone flyover of the coastline at golden hour"
first: <generated or uploaded image>
model: minimax/video-01

Story continuation

The node also emits the clip’s first and last frame as images, so a second Video node can start where the first ended (wire First frame / Last frame source handles into the next node’s first/last targets).

Notes

• Video generation is premium — a single clip can cost over €1, and the node card shows a badge to warn you.
• Generated clips are saved to the node’s artifact history.

Related

• Merge Videos — concatenate clips.
• Image Generation — producing start/end frames.