How much does it cost to make a real estate video from photos with AI?
About $1.11 for a 30-second video: five 6-second image-to-video clips on Seedance 1.0 Pro Fast at 720p cost $0.90 total, a generated music track costs a flat $0.20, and stitching the clips together with ffmpeg is free. A premium tier of the same model family costs $5.43 for the identical five clips.
Which AI model is best for animating property photos?
For slow camera moves over architecture, the budget tier holds up: Seedance 1.0 Pro Fast at $0.030 per second of output produces smooth push-ins and pull-outs at 720p. Seedance 2.0 Fast costs six times more per second and buys finer water and foliage detail plus native audio — which you rarely need when you are scoring the video with music anyway.
Can I mix portrait and landscape listing photos in one video?
Yes. The model keeps each photo’s own frame shape when aspect_ratio is left on adaptive, and at assembly time landscape clips are composited onto the vertical canvas over a blurred, darkened copy of themselves — nothing gets cropped away.
Generate a vertical video ad with AI for an iOS app: keyframe anchoring lets an image model own product identity and a video model own only the motion.
Image generation API pricing compared across 15 text-to-image models, from $0.0025 to $0.25 per image. Live per-model rates, pay-per-use, no subscription.
Get 4K, high-quality AI images without overpaying: generate on dedicated models, upscale only the keepers, and explore at low resolution to cut cost.
We took eight listing photos of a stone villa — pool, bedroom, jacuzzi terrace, night exterior — and turned them into the 30-second video below for $0.90 in model spend, plus $0.20 for the music. The recipe is short: animate each photo into a 6-second clip with an image-to-video model, generate an instrumental bed, and stitch it all together with ffmpeg. No footage, no drone, no editor.
The finished 30-second tour — five photos, five AI camera moves, $1.11 all-in.
The recipe
1
Pick five photos
Five photos at 6 seconds each makes a 30-second video. Mix wide exteriors with interiors, and order them day → interior → golden hour → night: the lighting arc reads as a story with zero editing effort. Keep every photo in its native orientation.
2
Animate each photo into a 6-second clip
One image-to-video call per photo. The photo becomes the first frame; the prompt describes only the motion.
3
Generate a music bed
One text-to-audio call, flat price, any length up to ~3 minutes.
4
Assemble with ffmpeg
Crossfades, fades, and the music duck — all free, all local.
The camera prompt that makes it look filmed
The single highest-leverage trick: describe the camera, not the property. The photo already fixes what everything looks like, so every word about the scene is wasted — and words about motion are not. This prompt pattern produced every shot above:
The prompt patternbash
Real-estate cinematography, one continuous take, no cuts.
The camera glides slowly and smoothly forward in a steady dolly
push-in toward the pool and the stone villa, gimbal-smooth and
perfectly level. The pool water ripples very gently, the leaves
sway lightly in a soft breeze. Warm midday sun, calm and luxurious
mood. No people, no text, no zoom snaps, no shaking.
Alternate push-in and pull-out across consecutive shots so the video breathes. Let one or two things move — water, curtains, leaves — and explicitly pin the rest (“the cushions stay perfectly still”). Then the call itself, here via the ModelRunner MCP or any HTTP client:
We rendered the identical five shots — same photos, same prompts — on two tiers of the same model family. Prices below are read live from the catalog; the totals are what our runs actually billed.
The same 30-second video on both tiers. Live prices show each tier's lowest (480p) per-second rate; the five-shot totals are what our 720p runs actually billed, at $0.030/s and $0.181/s.
Same source photo, one clip from each tier — judge the 6× gap yourself:
Seedance 1.0 Pro Fast — this clip billed $0.18.Seedance 2.0 Fast — this clip billed $1.09.
For slow camera moves over architecture — no faces, no fast motion — the cheap tier holds up remarkably well. The premium tier renders finer water shimmer and foliage detail, and generates synchronized ambient audio. But if you are scoring the video with music, that audio gets muted anyway; do not pay 6× for a feature you will discard. Two useful knobs on the budget tier: resolution: "480p" halves the rate to $0.015/s, so drafting all five moves costs $0.45 before you commit to finals, and leaving aspect_ratio unset keeps each photo’s own frame shape.
Music and assembly
Stable Audio 2.5 bills a flat $0.2 per audio clip per generation whether you ask for 10 seconds or 3 minutes. Prompt it like a music brief: name the genre, tempo, and mood, and say “fully instrumental, no vocals” for a clean bed. Then ffmpeg stitches everything — 0.5-second crossfades between clips, a fade at both ends, music underneath:
Assemble three clips + music (extend the pattern for five)bash
Mixed a landscape photo into a vertical video? Composite it over a blurred, darkened copy of itself instead of cropping — the full frame survives and the canvas stays filled.
What it all cost
$1.11
Budget route: 5 clips + music
$5.63
Premium route: same video on Seedance 2.0
$0.45
480p draft pass of all five shots
At scale the marginal cost per property is just the video spend — the assembly script and the prompt pattern are reusable across every listing. A hundred properties on the budget tier is $111, not a videographer’s day rate.
