Grok Imagine first and last frame: keyframes on xAI only

xAI takes up to 4 timestamped keyframes for Grok Imagine video. Sume's grok-imagine-video-1.5 row takes one start image and rejects an end frame.

4 min readSume
All posts

xAI's Grok Imagine docs describe keyframes: up to 4 {image, timestamp_s} entries. The Sume Grok row has end_frame: false, takes one start image, and lists "no end_frame" in its constraints.

xAI's keyframes are from its video docs, read 2026-09-30. Sume's rules are from its catalog and Video Router docs.

What does the Sume Grok row accept?

One start image, as image_url or first_frame_url, and nothing at the end. There is no timestamp-based keyframe list on that row.

Which Sume ids take an end frame?

The Video Router docs show gemini-omni-flash-1.1 accepting image_url plus an optional end_image_url. The catalog marks end_frame: true for wan-3.0, and MiniMax H3 Max is documented with first/last-frame image-to-video.

From xAI's docs and Sume's catalog and docs, read 2026-09-30: start and end frames.
WhereFrame inputs
xAI, Grok ImagineUp to 4 {image, timestamp_s} keyframes
Sume, GrokStart image only
Sume, gemini-omni-flash-1.1image_url + optional end_image_url
Sume, wan-3.0end_frame: true in the catalog

Where can I read more on first and last frames?

See Kling 3 first and last frame and first and last frame image-to-video for ids that take both.

What should I do?

If you depend on mid-clip keyframes, use xAI directly. On Sume, pick an id whose catalog row lists end_frame and send a start and end image.

Sources

Related posts

More in Developers

All Developers posts

Written by Sume