AI video for jewellery: animate a real jewelry photo

To make an AI video for jewellery (jewelry in American spelling), start from a real product photo as the video's first frame, ask for one slow move such as a turn or a light sweep across the stones, keep the clip short, and check the stones and engraving frame by frame before you publish. To make it loop, use the same photo as the last frame too.

The request below uses Sume's Video generation and Video frames docs, read on 2026-09-28.

How do I make a jewellery video from a photo?

Send the photo in frame_images on POST /v1/videos. An entry with frame_type first_frame makes the request image-to-video, so the clip opens on your real piece; a last_frame entry with the same photo brings it back there. Pick a model whose supported_frame_images lists both, and send resolution explicitly.

curl -X POST "https://api.sume.com/v1/videos" \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2",
    "prompt": "Slow light sweep from left to right across the diamond ring on a dark velvet stand. The ring stays still; only the light moves. Soft sparkle on the stones, no camera shake.",
    "frame_images": [
      { "type": "image_url", "image_url": { "url": "https://example.com/ring.jpg" }, "frame_type": "first_frame" },
      { "type": "image_url", "image_url": { "url": "https://example.com/ring.jpg" }, "frame_type": "last_frame" }
    ],
    "aspect_ratio": "9:16",
    "resolution": "1080p",
    "duration": 5,
    "generate_audio": false
  }'

Which moves suit jewellery?

One move per clip. The docs advise detailed prompts that cover motion, camera angles, lighting, and composition. Moves that keep the piece readable:

• A light sweep: the piece stays still and a highlight travels across the metal and stones.
• A slow push-in toward the main stone.
• A partial turn on a stand. A full turn from one photo makes the model invent the back; AI 360 rotation video pins front and back photos instead.
• A small wrist or head turn on a model, starting from an approved on-model still. AI model for jewelry covers that still.

Which settings should I send?

Most catalog models top out at 15 seconds. Among the Seedance ids, seedance-2 and seedance-2.5 offer 1080p.

How do I check the stones and engraving?

Pull stills from the finished clip with POST /v1/video-frames. It reads a clip already on Sume, this workspace's media.sume.com file, takes at[] with 1–24 times in seconds, and returns durable images; ask for png for lossless inspection. Frame extraction is unbilled. Compare the stills with your photo: stone count, prongs, chain links, and any engraving. The first and last frames are pinned, but everything in between is generated. Product logo warping in image-to-video explains the difference between a pinned frame and a reference, which only guides.

Is there a ready-made recipe, and what does it cost?

The catalog Format sume-cinematic-studio-commercial describes itself as a “cinematic studio commercial with premium lighting, macro product detail” and lists “luxury objects” among its uses; AI product commercial video API shows how to call a commercial Format. A direct /v1/videos job is reserved on submit at the provider's list price × 1.25, plus a 5.5% agent fee by default, and usage.cost on the job is the billed amount.