Reimagine a Short from one frame: build your own 8-second clip

YouTube's Reimagine takes one frame from an eligible Short and turns it into a new clip eight seconds long; on Sume you can build the same shape from footage you own by extracting that frame with video frames and sending it to Gemini Omni Flash 1.1 as an image-to-video request, which takes 3 to 10 seconds. The difference is control: Reimagine runs inside YouTube on other people's eligible Shorts, while this route runs on your own files and gives you the file back.

YouTube's side comes from its Reimagine announcement, read on 2026-10-03. Sume's side comes from the docs for video frames and the Video Router.

What does Reimagine do?

Per YouTube's post, published on March 18, 2026, Reimagine is powered by Veo and appears under the Remix tool on eligible Shorts. It makes an 8-second clip from a single frame. A creator can insert themselves or objects using up to two references from their photo gallery, and prompt with an AI-suggested or custom text prompt to get video with audio. Every Reimagined Short links back to the original work.

Can I do this with my own Short on Sume?

Yes, as a different tool for a different job. Sume cannot run on YouTube's eligible-Short pool, and it is not Veo. It can take your own clip, give you a still, and animate that still. The Video Router lists gemini-omni-flash-1.1, which accepts image_url for image-to-video, 3 to 10 seconds, 16:9 or 9:16, at 360p, 720p, 1080p or 4K, with synced audio always on.

Start by importing the clip so it lives on media.sume.com; the docs say there is no open-internet fetch. Then call video frames with at set to the second you want, for example [2.5]. The frame comes back as a durable image at source size, or clamp it with max_edge between 16 and 2160. At most 24 frames per call, source up to 300 seconds.

What does the animate step look like?

Send the frame as image_url with a prompt that says what should happen next, duration 8 to match Reimagine's length, and aspect_ratio 9:16.

If you want yourself or an object in the new clip, the same model takes reference_image_urls (up to 10) in a reference-to-video request, addressed in the prompt as <IMAGE_REF_0>. Sume routes the one model id by the shape of the request, so send either image_url or the reference lists; the edit field video_url cannot be combined with either.

curl -X POST https://api.sume.com/v1/video-frames \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: reimagine-frame-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/short.mp4",
    "at": [2.5],
    "format": "png"
  }'

What does it cost?

Video frames is billed by its own Modal compute, so check the job rather than assuming a flat rate. Gemini Omni is billed at provider list times 1.25 per output second. At 720p the provider list is $0.10 a second, so an 8-second clip is $1.00 on Sume; at 1080p, $0.15 list, it is $1.50.

What Sume does and does not do here

Sume does not edit inside YouTube, does not use Veo, and does not add the link-back that Reimagine adds. If you publish a clip built from someone else's frame, the credit decision is yours, and the remix opt-out rules sit on YouTube's side.

• Does: extract a frame from a clip you own, animate it, return a file you can edit further in Timeline.
• Does: keep native Omni audio, or replace it by putting your own audio spine on a Timeline render.
• Does not: remix other creators' Shorts on YouTube's behalf.
• Does not: guarantee the output matches your frame beyond what the prompt and model deliver; check the result before you post.