AI avatar for website videos: a talking welcome clip

An AI avatar for a website is a short presenter video on the page: a generated person who says your welcome, product intro, or about-us message inside the page's video player. On Sume it is a rendered MP4, not a live chat character. Render the script as an Avatar 1.0 talking video in the shape of the page slot, burn in captions for visitors who watch with the sound off, and embed the file with an HTML video element.

Sume facts come from the Generate avatar video and Avatar video previews docs, read on 2026-09-28; anything called current behavior is read from Sume's code. Browser behavior comes from MDN's video element reference and autoplay guide.

Which shape should a website avatar video be?

Pick aspect_ratio from the slot the video fills. Avatar 1.0 renders five shapes, and 720p is the documented resolution. The API reference says a value outside the supported set is treated as the default, 9:16, rather than refused, so double-check the ratio you send; talking head video format covers the five shapes and cropping to others.

How do I make a website welcome video with an AI avatar?

Use the preview route, so you approve the first frame and set captions in one flow:

• Create or pick the presenter once; How to create a reusable AI avatar covers that request.
• Keep the script short. Sume accepts an estimated 4–60 seconds per video.
• Create the preview with captions, as below. Preview stills are never captioned; the stored captions are burned in when you call POST /v1/avatar-video-previews/{id}/generate-video. In current code, POST /v1/avatar-1.0/talking-video refuses a captions field.
• Read the result with GET /v1/avatar-videos/{id}: the video_url for the player and the preview_image_url still for the poster.

curl -X POST https://api.sume.com/v1/avatar-video-previews \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: homepage-welcome-v1" \
  -d '{
    "avatar_handle": "site_host",
    "script": "Welcome to Acme. Here is what we make, who it is for, and where to start.",
    "scene": { "type": "prompt", "prompt": "Bright modern office, plants, soft daylight" },
    "aspect_ratio": "16:9",
    "captions": { "enabled": true, "style": "slam" }
  }'

Why should a website avatar video have captions?

Because a welcome clip that starts by itself should start muted. MDN's video reference says modern browsers block audio, or videos with an unmuted audio track, from autoplaying. Its autoplay guide says that blocking doesn't apply to a muted track, and lists a visitor's earlier interaction with the site and an allowlisted site among the other conditions that allow autoplay. Burned-in captions carry the message until a visitor turns the sound on. Two caption rules to plan for:

• Inline captions refuse an estimated duration above 60 seconds.
• A caption failure doesn't fail the video: the job can still succeed with a clean video_url and captions.status=failed, so check that field before you publish.

How do I put the avatar video on my page?

Use an HTML <video> element with the finished file as src and the preview still as poster, which MDN describes as the image shown while the video is downloading. For autoplay, MDN's guide pairs autoplay with muted and with playsinline, which it says Safari requires for autoplay; controls lets visitors unmute.

• The player can point at the media.sume.com URL, which the docs call public. For a page you'll keep, copy the MP4 and the still to your own storage or CDN; Do AI-generated video URLs expire? explains which Sume URLs the docs call durable.
• Make the video from your server. Sume's docs say not to place API keys in frontend JavaScript, and the page only needs the finished file.

<video
  src="https://example.com/media/welcome.mp4"
  poster="https://example.com/media/welcome-poster.png"
  controls
  autoplay
  muted
  playsinline
></video>

What can't a website avatar do?

• Talk with visitors. Each video is a job you submit, poll, and read, and the page plays the finished file.
• Speak other languages in the same request: in current code, Avatar 1.0 speech is English only. Other languages go through TTS 1.0 and Fabric lip sync.
• Run past 60 seconds in one video. Split a longer script and join the parts, as in How to make an AI avatar video longer than 60 seconds, and caption each part first.

What does a website avatar video cost?

Avatar video bills per second by quality tier: $0.184/s standard, $0.245/s plus, $0.55/s max (no product image). A 30-second welcome at the default plus tier is $7.35 of avatar video. Inline captions are a separate add-on on the avatar-video estimate, and creating an avatar costs $0.95 per avatar. Each is plus a 5.5% agent fee by default.