One avatar handle, three platform cuts: keep the character consistent

To keep one character across LinkedIn, Snapchat and Pinterest, create the avatar once and reference the same avatar_handle in every render. Only the ratio, length and script change per platform. Sume stores the handle without a leading @, so the same presenter comes back every time.

How does the avatar stay the same?

The Create new avatar docs describe three inputs: a prompt, a profile of traits (props) or a reference image (photo). Each request creates a job, and you use the returned handle or resource id for videos. Avatar video requests reference the avatar with a top-level avatar_handle. So the consistency is structural: the same stored avatar, not a fresh prompt per clip.

The honest caveat: the docs say current execution supports one resolved avatar per final video, and scene backgrounds resolve to one shared scene. For a cut with two presenters you need two separate renders.

What changes per platform?

Length and shape change. Here is what the vendor pages I fetched say next to what Sume offers.

How do you check a look before three renders?

Use a preview. The Avatar video preview route makes the first-frame still without starting the full render. The docs say you can regenerate stills and, once happy, call generate-video on the preview. Preview stills are never caption-burned, and for multi-scene previews later stills are pose-anchored continuations of the first frame.

• Create the avatar once and note the handle.
• Preview the first frame at 9:16 and 1:1, approve the look.
• Render each platform's cut with its own script length and Idempotency-Key.
• Probe each finished file for duration before you upload it.

Request shape for a LinkedIn square and a Snapchat vertical

Same handle, different aspect_ratio. The default is 9:16, and resolution is currently 720p.

curl -X POST https://api.sume.com/v1/avatar-1.0/talking-video \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: host-linkedin-001" \
  -d '{"avatar_handle": "product_host", "aspect_ratio": "1:1",
       "script": "A short tip for busy product teams."}'

What Sume does not promise

Sume does not claim pixel-identical faces across renders, and I found no such guarantee in the docs. The handle gives you the same stored avatar each time; lighting and framing come from each scene prompt, so reuse the same scene text when you want the same room. Publishing to any of the three platforms is outside Sume, as covered in the earlier posts on consistent characters in AI video.