Real estate walkthrough video from listing photos: the 30-image limit
A Sume Format run accepts up to 30 images as attachments. How to send listing photos, keep room order, set a cap and read the walkthrough back.

You can build a real estate walkthrough video from listing photos by sending the photos as attachments on a Format run. The create body accepts up to 30 images in attachments ({ "type": "input_image", "image_url": "https://..." }), and the whole run shares one media budget of 30 files with a maximum of 10 videos and 10 audio files. A listing with 42 photos therefore needs a choice: send the best 30, or split the listing across several runs.
Sending the photos
Put the room order in the instruction, because the agent reads attachments as images and not as a labelled list. Put structured facts such as the address, the bedroom count and the asking price in input. Sume saves input as a data file and tells the agent it is data, so listing text scraped from a portal does not get read as commands.
For a first look at what a catalog Format expects, read it before you call it. The sume-slideshow slug is listed in the catalog at the sume handle; the io block on the response shows its declared input and output kinds.
curl -sS "https://api.sume.com/v1/formats/sume/sume-slideshow" -H "Authorization: Bearer $SUME_API_KEY"The limits that matter for a listing
All items in the table come from the call docs, as of 2026-10-09.
| Limit | Value | Effect on a listing |
|---|---|---|
Images in attachments | 30 | Pick the best 30 photos |
| Media files in one run | 30 total, 10 videos, 10 audio | Counts URLs inside input too |
input size | 64 top-level keys, 2 MiB | Room notes fit easily |
instruction accepted | 8000 characters | Only about 4000 are carried as prompt text |
generation_spend_cap_usd | Up to 500; default from the Format | Set about 20 for a short tour |
Reading the result
When the run is completed, primary_output_url is the video, a durable media.sume.com URL that never expires. Anyone who has the URL can open it, so a listing video meant for one buyer should be copied behind your own access control. If you want the length to be checked before posting, read duration_ms on the artifact entry.
Keep your listing id on your side. It cannot come back in output unless the run repeats it, so store the run id next to it. Use the listing id and a version as the Idempotency-Key, so a retry after a network error does not start a second paid run.
Listing photos: practical points
Send photos that your agency owns or has the right to use, and send only HTTPS URLs that stay reachable for the length of the run. An unknown asset_id returns 400 attachment_not_found, and an attachment over the media budget returns 400 invalid_attachment, so the problem shows up at create and not after a paid run.
Order matters to a human viewer even though the API does not ask for it. Number the photos in the instruction, for example "photo 1 is the street view, photos 2 to 5 are the living room". If the first result skips a room, continue the run with previous_run_id and name only the room that needs fixing, so you do not pay for the whole tour again. Do not describe features the photos do not show; the agent follows your text, and a tour that claims a garden that is not there is a problem for the listing, not for the model.
Sources
Related posts
More in Use cases
- Vertical recipe short on Wan 3.0: 16 s with a plated-dish last frame
A 16-second 9:16 Wan 3.0 recipe clip costs $1.00 at 480p, $2.00 at 720p and $4.00 at 1080p on Sume, and the vertical ratio does not change the price.
- Reels ads: sound strongly recommended, captions optional; 12 for $2.40
Meta: captions optional but recommended; sound optional, strongly recommended on Reels. Captioning 12 thirty-second ads on Sume is $2.40 per the docs.
- Swap a spokesperson in a 14-second clip: face swap beta $2.58 to $7.70
Sume's Avatar Face Swap beta prices a 14-second regional ad clip at $2.58 standard, $3.43 plus or $7.70 max. Source must be 4 to 15 seconds with usable audio.
- Restaurant daily clip for 30 days: 5 seconds on Wan 3.0, $9.38 at 480p
A new 5-second restaurant clip every day for 30 days is 150 seconds of video: $9.38 at Wan 3.0 480p, $18.75 at 720p, $37.50 at 1080p on Sume.
Written by Sume