Video prospecting with an AI avatar: one clip per prospect

Video prospecting is sending a short, personal video inside a sales outreach message, such as an email or a social message, to start a conversation with someone who doesn't know you yet. The video names the prospect, says why you are reaching out, and asks for one next step.

Recording each video by hand limits how many you can send. With an avatar API you write one script template, fill in each prospect's name and company in your code, and render one short video per prospect. On Sume that is one talking-video job per prospect, each with its own idempotency key and webhook, paced by your plan's concurrency. Facts come from Generate avatar video and Generation admission, read on 2026-09-28, plus Sume's current code where noted.

What makes a good prospecting video?

• The prospect's name and company in the first sentence, so the video is plainly for them.
• One reason for reaching out, tied to something about their company.
• One ask: a reply, or a short call.
• Short. One Sume avatar job covers an estimated 4-60 seconds, and the outreach script can be much shorter than that.
• A still as the link image. Completed results can include a preview_image_url next to the media.sume.com video.

How do I make one video per prospect?

Loop over your prospect list in code. For each row, fill the template, then submit one talking-video job. Store the job id with the row, so a crashed script can pick up where it stopped instead of submitting again, and keep polling status_url as a backup for missed webhooks.

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: outreach-v2-prospect-0042" \
  -d '{
    "avatar_handle": "outreach_host",
    "script": "Hi Jane, I saw Acme just opened a second office. Worth a short call?",
    "aspect_ratio": "9:16",
    "webhook_url": "https://example.com/hooks/sume"
  }'

How many prospect videos can I render at once?

As many as your plan's processing concurrency allows. Extra valid jobs are accepted as queued while queue capacity remains, and a new submit fails with 429 queue_full only when the queue is full too. Video job concurrency and queueing lists the per-plan numbers and how to read your workspace's effective limit.

Will the avatar say each name correctly?

It reads every name as English: in current code the avatar route speaks English only. Check a few unusual names before a large send. Personalized video at scale covers saying names in a batch.

Should the avatar look like me?

A prospect may expect the sender, so decide this before you send. An avatar made from your photo is a generated likeness: in current code a photo input is redrawn by an image model prompted with "Photo of this person", and the avatar's voice is cloned from a sample clip Sume generates from the avatar, not from a recording of you. Cloning yourself with AI covers a route that uses your own voice.

What does a prospecting video cost?

Avatar video bills per second at $0.184/s standard, $0.245/s plus, $0.55/s max (no product image), plus a 5.5% agent fee by default; see API pricing. AI avatar video pricing per minute works through what else changes the bill.