Real-estate listing video: 30-second Wan 3.0 clip from ten photos
Turn ten listing photos into a 30 second video on Sume: one reference-to-video clip or ten 3 second photo clips, both $3.75 at 720p. Which is more faithful.

You can make a 30 second listing video from ten photos on Sume two ways, and both cost $3.75 at 720p: one wan-3.0 reference-to-video request with ten images, or ten 3 second clips that each start from one photo. The second is more faithful to the property, because each photo is a real first frame.
Reference-to-video: one request
Sume's Wan 3.0 accepts up to 10 reference images on one request, and the docs say references are visual guidance, not exact frames. A model given ten photos of a house may blend, simplify or invent rooms. For a listing, that is a risk: a video that shows a room that does not exist is a problem for the agent and the buyer.
Image-to-video: ten requests
Image-to-video starts from your photo as the first frame, so the opening of each clip is the real room. Wan 3.0 accepts 2 to 30 seconds, so ten 3 second clips make 30 seconds. You then join them in an editor. The per-second price is the same, so the extra cost is the joining step, not the render.
What it costs
Both routes charge for output seconds, so the table is the same for either method at a given resolution.
| Resolution | Per second | One 30 s request | Ten 3 s clips |
|---|---|---|---|
| 480p | $0.0625 | $1.875 | $1.875 |
| 720p | $0.125 | $3.75 | $3.75 |
| 1080p | $0.25 | $7.50 | $7.50 |
Per month
For twenty listings a month at 720p, that is 20 x $3.75 = $75.00; at 480p drafts for the agent to approve, 20 x $1.875 = $37.50.
Prompt for faithful motion
Prompt for motion that does not change the structure: a slow push-in, a pan across the room, a gentle dolly. Avoid words that ask the model to add things ("add a sofa", "make it brighter") unless the listing will say the photo is virtually staged. Keep each prompt short and name the room.
curl -X POST https://api.sume.com/v1/videos \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: listing-12-kitchen-v1" \
-d '{
"model": "wan-3.0",
"prompt": "Slow push-in on the kitchen, natural light, no people",
"resolution": "720p",
"duration": 3,
"aspect_ratio": "16:9",
"frame_images": [{"type": "image_url",
"image_url": {"url": "https://example.com/kitchen.jpg"},
"frame_type": "first_frame"}]
}'Sound
Audio is on by default for Wan 3.0 on Sume (generate_audio defaults to true in the code). A property video usually carries a music bed or a voice-over from your editor, so consider switching generated audio off with generate_audio: false and adding your own track in post. Check the field in the Video generation docs for the current default before relying on it.
Review and disclosure
Check each clip against its photo before publishing: wall colors, fixtures, window positions. Disclose generated motion if your market or your listing platform requires it; we did not verify any platform's rules here, so read theirs. For a vertical cut, use aspect_ratio: 9:16 from the start and prepare the photos for that crop.
A shooting and review sequence
A practical sequence for a listing: shoot or collect ten photos in the order of a walk-through (exterior, entry, living room, kitchen, dining, primary bedroom, bathroom, second bedroom, outdoor space, exterior at dusk). Run a 3 second 480p clip from each, which costs 3 x $0.0625 = $0.1875 per photo and $1.875 for the set. Review them, regenerate any that distort the room, and only then render the finals at 720p or 1080p.
The draft set costs less than a single 1080p clip of five seconds ($1.25). That budget logic is why the first pass should always be the cheapest tier: a bad room found at 480p costs under 20 cents, and the same fault at 1080p costs 75 cents.
Joining the clips
Join the ten clips in a timeline or video editor; any editor that accepts MP4 files works. Keep a consistent aspect ratio across the requests, since each clip is a separate job. For a vertical listing reel, send aspect_ratio: 9:16 on every request and prepare the first-frame photos for that crop.
Vendor versus Sume limits
Wan 3.0 is Alibaba's model, and its vendor page lists reference-based generation up to 20 assets (read 2026-10-05); the Sume request is narrower, at 10 images. The Sume request fields come from the Video Router docs and Video generation docs. Source for the vendor claim: the Wan 3.0 repository.
Sources
Related posts
More in Use cases
- Realistic AI b-roll in a Short: is the Altered content flag needed?
A realistic generated scene that did not occur needs YouTube's disclosure; a fantasy scene does not. Sort Sume clips before upload using the page's own list.
- Recipe blog post to video: scrape text, send step photos to Wan 3.0
Wan 3.0 can take a webpage as a reference at Alibaba, but Sume does not expose web_url. Scrape the recipe, then send step photos as references. Steps and curl.
- Recolor a product with a swatch reference and mask on GPT Image 2.5
BFL's FLUX 3 Image lists recolor, replace and move edits by bounding box. On Sume, a recolor job takes a swatch reference and a mask_url on GPT Image 2.5.
- Redact names from a transcript and the audio: STT words, then split
Sume STT returns word timings with a type field. Find the names you list, blank them in text, and cut the same spans from the audio for $0.01.
Written by Sume