Real-time or async avatar: six questions to ask before you choose

Choose real-time only when a person must talk with the avatar and get an answer back in the same moment. For everything else, a finished clip is simpler and cheaper to run. Sume avatars are async jobs: you submit a request, get a job id back, and read the finished video later. There is no live video session. This post is a decision checklist, not a comparison of live products. Answer the six questions below and count how many say "live".

The interest in live conversational video has grown, so teams often start by asking for real-time and later discover they wanted a good spokesperson clip. The questions separate the two.

The six questions

What a Sume avatar clip gives you

A Sume Avatar 1.0 talking video takes a ready avatar handle and a script, or ordered video_inputs scenes. The estimated duration must be between 4 and 60 seconds. You pick quality (standard, plus, max), an aspect ratio (1:1, 3:4, 9:16, 4:3, 16:9), and resolution is 720p at this time. You can ask for burned-in captions after generation.

Because the clip is a file, you can review the script, check the first frame with an Avatar video preview, and publish the same video in many places. Cost follows seconds, not viewers: at the plus no-product rate of $0.245 per second, a 30-second clip is $7.35 no matter how many people watch it.

What it does not give you

It does not give you a live video session, a microphone input, or an avatar that hears the viewer. The job model is submit, wait, fetch. The wait is a queue plus a render, and the docs say video and avatar-video jobs usually take longer than the 30-second wait that sync mode can hold open, so build for polling or a webhook.

A mixed design

Many teams do both. They use chat, voice, or a human for the live part, and use Sume to make the prepared parts: the welcome clip, the explainer for each product, and the recap after a call. Prepared clips carry a consistent face and script, and the live part handles the unpredictable questions.

• Zero or one "live" answers: use async clips.
• Two or three: use async clips for the prepared parts and something live for the rest.
• Four or more: you need a live product; Sume does not provide one.

Worked example: a product launch

Take a launch week. You want a presenter on the landing page, a short clip in each customer email, and a recap video that goes out after each live Q&A. The landing page message is the same for every viewer. The email clips are per-segment. The recap is written after the Q&A and approved by a human before it goes out. Four of the six questions answer "async" without any trouble. The Q&A itself is live, but it is a human with a microphone, not an avatar.

Cost for the prepared parts: one 30-second landing clip at plus is $7.35, five 20-second segment clips are $24.50, and one 45-second recap is $11.025. The total is $42.875 at no-product rates, a figure you can approve before you start. A live avatar would bill by the minute of conversation and would depend on how long people stay, which is much harder to budget.

The practical rule: use live when the conversation is the product, and async when the message is the product.