Edit a video with a text prompt: Grok Imagine vs Sume Omni

If you want to edit an existing video with a text prompt, xAI documents that for Grok Imagine. Sume's grok-imagine-video-1.5 row is image-to-video only, so on Sume the edit route is gemini-omni-flash-1.1 with a video_url. For swapping the people in a clip there is also h3-max-recast.

Edit options side by side

The comparison below separates what is documented on each side.

The Sume edit request

The Sume edit body is small. video_url is the source clip, prompt says what to change, and resolution picks the output tier. Sume sends no aspect_ratio for edits (it returns 400 if you give one) and no duration to the provider. You cannot combine video_url with image_url, end_image_url or reference_*_urls.

generate_audio: false, bitrate_mode and reference_audio_urls are rejected for this model: it always produces native synced audio.

{
  "model": "gemini-omni-flash-1.1",
  "prompt": "Replace the bottle with an apple. Keep everything else the same.",
  "video_url": "https://example.com/clip.mp4",
  "resolution": "720p"
}

Price and route

An edit bills output seconds. A duration sent with video_url is only a reserve hint, and the default hint is 8 s. At 720p Sume lists Omni at $0.125 per second (fal list $0.10 x 1.25), so the default 8 s hold is 8 x $0.125 = $1.00.

The edit route is a Video Router field, not part of the OpenRouter-shaped /v1/videos body. Send it to POST /v1/video-router/generate. The docs say sume/auto can select an edit-capable family for video_url requests, and name the Omni 1.1 fallback.

Bottom line

Grok Imagine editing is a vendor feature that is not in the Sume catalog. If your pipeline needs it, call xAI directly. If you only need a prompt-driven edit with sound, use the Omni route above.