Tavus Griffin single photo input vs a Sume photo avatar handle

Yes, Sume can start from a single photo too, but it works differently from Griffin. Tavus says Griffin generates a live, conversational video from one reference photograph, and the model is a research preview for select testers. In Sume, one photo becomes a reusable avatar through POST /v1/avatar-1.0/generate for $0.95 flat, and every later clip references that avatar by handle.

Create the avatar from a photo

The creation request takes avatar_handle and an input union. The three forms are a prompt, props (ethnicity, sex and age) or a photo object with an image_url. A leading @ in the handle is stored without it.

Media inputs must be public HTTPS URLs. The media input rules reject localhost, private-network, signed or private URLs and mismatched content types before submission, so host the photo somewhere public and with the right content type.

curl -X POST https://api.sume.com/v1/avatar-1.0/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "avatar_handle": "product_host",
    "input": {
      "photo": { "image_url": "https://example.com/host.jpg" }
    }
  }'

What each side asks of you

Tavus's page frames the single photo as the entire input to a live model. The Sume flow adds an explicit step, which costs once and pays back across clips: one handle, many scripts, one consistent face.

Then render with the handle

Once the avatar is ready, an avatar video request names the handle and gives exactly one of script or video_inputs. Pricing per second depends on quality and whether a product_image is supplied. At standard, the rate without a product is $0.184 per second, so a 10 second clip is $1.84 on the Sume API pricing page.

Preview the first frame before you commit to a full render if the photo is a close call.