Midjourney V8.2 edit stills, animated on Sume

Edit your still in Midjourney, put the final file at a public HTTPS URL, and send it to Sume as the first frame of a video request. On POST /v1/videos that is a frame_images entry with frame_type: first_frame; on the legacy Video 1.0 URL it is image_url.

What the Midjourney notes say

The Midjourney notes on Releasebot list a Sep 3, 2026 entry for a V8.2 edit model in the lightbox, with plain-instruction edits using up to 4 reference images. An Aug 27 entry describes the first V8.2 edit model with inpaint and outpaint, and a Sep 24 entry adds Korean language on midjourney.com. The Sume docs do not list a Midjourney model id, and this post does not cover generating inside Midjourney.

From still to first frame

Image inputs on Sume must be fetchable public HTTPS URLs. Localhost, private-network, non-HTTPS and signed or private URLs are rejected, as are mismatched content types. So download the Midjourney result and host it where Sume can fetch it.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: mj-still-anim-001" \
  -d '{
    "model": "sume/auto",
    "prompt": "Slow push-in; keep the subject sharp and the background drifting",
    "aspect_ratio": "9:16",
    "duration": 5,
    "frame_images": [{
      "type": "image_url",
      "image_url": {"url": "https://example.com/final-still.png"},
      "frame_type": "first_frame"
    }]
  }'

Edit with references, then keep them

If you edited with several reference images in Midjourney, you can reuse that idea on Sume. input_references gives reference-to-video guidance rather than exact frames, and only models whose supported_input_references lists a type accept it. If both frame_images and input_references are sent, frame_images takes precedence.

Poll the polling URL the submit returns, or GET /v1/jobs/{id}/status. Auto create controls default to 720p and 8 seconds, with 3 to 10 second clips at 16:9 or 9:16. Check GET /v1/videos/models for the exact limits of a model you pin.