Airbnb host welcome video with an AI avatar in your own living room

A short welcome clip answers the same five questions every guest asks: where to park, how to get in, the wifi, quiet hours and who to call. You can have an avatar say it, placed in a photo of your actual living room, so guests see the space they booked. This post works out the cost of a 30 second version and the request that makes it.

The scene is your photo

The talking-video route accepts scene: { "type": "photo", "image_url": <url> }, a public HTTPS image used as the scene reference. Point it at a well-lit shot of the room. Scene direction can also be a text prompt, but a photo keeps the clip honest about the place. Create the avatar once ($0.95, flat) and reuse its handle for every listing.

What 30 seconds costs

Thirty seconds is roughly 75 spoken words, enough for check-in, wifi and one house rule. Sume bills per second of finished video, and these rates apply with no product image.

The request

Use 9:16 if guests get the clip by text message and 16:9 if you embed it in a guidebook page.

curl -X POST https://api.sume.com/v1/avatar-1.0/talking-video \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"avatar_handle":"@host","quality":"plus","aspect_ratio":"9:16","scene":{"type":"photo","image_url":"https://example.com/living-room.jpg"},"script":"Welcome! Parking is on the left of the driveway. The key box code is in your booking message. Wifi is on the fridge. Quiet hours start at ten. Text me anytime."}'

Check before you pay for the top tier

Generate a preview still first with POST /v1/avatar-video-previews. Previews do not depend on quality tier, so after you approve the still you can choose standard to save money or max for the final version without re-previewing. Re-record the clip when the door code or wifi changes; at standard that is $5.52 per update.

What not to say

Skip anything that changes often. Door codes and wifi passwords belong in the booking message, not in a video you pay to re-render. Say where to find them instead. Keep the avatar's tone like a host, not a hotel: one greeting, three facts, one invitation to text. That keeps the script near 70 words and the cost near $7 at the default tier.

If you manage several listings, create one avatar and reuse it, and swap only the scene.image_url and the script per property. Ten listings at 30 seconds cost 10 x $5.52 = $55.20 at standard.