Video models
The video models behind Crust's video nodes.
Video models offer some of these modes: Generate (a clip from a prompt, with or without start and end frames), Reference (images, clips or sound steer the result), Edit (change a clip) and Extend (continue one). Each card lists the modes a model offers, whether it makes sound, and how many references it takes. Durations and resolutions are on each model's page.
Video model families
veo-3.1-generate-001 / fastGenerate and reference modes. Start/end frames, up to three references, and optional generated audio. Extend is retired.
Seedance familyText, image, and reference video generation with large image/video/audio reference banks and optional generated audio.
standard / proGenerate mode with start/end frames and optional generated audio.
wan3.0-video / primeGenerate, reference, edit, and extend with image, video, and audio references.
grok-imagine-video / 1.5Generate, reference, edit, and extend. Supports a start image, source video, and several image references.
h3-maxMultimodal generation from prompt plus image, video, and audio references.
ray-3.2Text-to-video, image-to-video, video edit, start/end frames, and keyframe direction.
auto / v1.1Text, image, reference, and video-edit modes with dialogue-focused output.
Wan VACE / Luma Ray 3.2Generates or rebuilds a clip for a target aspect rather than only cropping it.
Sync V2 Pro / HeyGen Speed / PrecisionDrives mouth motion in a source video from the connected audio track.
ltx-2.5-proText-to-video or image-to-video with start and end frames, 6, 8, or 10 seconds, camera motion presets, and generated audio. Also Audio to video's Auto choice for 2–10 seconds of audio.
ltx-2.3-qualityAudio to video for longer audio, up to 60 seconds: a clip timed to the connected track.
omnihuman-v2A lip-synced talking person from one portrait and 4 seconds to 5 minutes of speech. Prompt optional.
sam-3-video-rlePrompted video masks; returns a mask track for the described subject.
proteus / artemis / nyx / gaia / starlightVideo upscaling with a choice of enhancement models; Auto picks by source.
meshy-7.1Text, image or multi-image to a GLB model.
Guide version 2026.10.21-3 · October 21, 2026