Cost per viewer of a 20 second avatar clip at 10, 100 and 1,000 views

A rendered avatar clip is a one-time cost, so its price per viewer falls as viewers rise. A 20 second Sume clip on standard costs 20 x 0.184 = $3.68. Shared across 10 viewers that is $0.368 each, across 100 it is $0.0368, and across 1,000 it is $0.00368. A live session is different, because each conversation uses the model for its own duration.

What this ignores

The arithmetic counts generation only. It leaves out hosting, bandwidth, your editing time, and the cost of re-rendering when the message changes. Sume serves results from public media.sume.com artifacts, so you do not need to host the file to preview it, but a production page will probably copy it to your own delivery path.

It also assumes one version for everyone. A personalized clip per recipient is a separate render for each, at the full $3.68 each on standard, so the viewer count in the table does not apply.

The live comparison, without invented prices

Tavus lists Griffin-Lite as a research preview for select trusted testers and not as a product for customers, and its page, as we read it on 2026-10-08, states no price. So there is no live price to put next to the table. When one is published, divide the per-minute price by the minutes of conversation each viewer actually uses; that is the figure to compare with the rendered cost per viewer.

When the number matters

Rendered clips win on cost when many people need the same answer: an onboarding step, a pricing explanation, a product launch. They lose when each person needs a different answer, since you either render per person or send them to a live agent or a person.

A good test is to count how many different questions your last hundred visitors asked. If twenty questions cover most of them, ten to twenty clips at about $3.68 on standard cost $37 to $74 in total.

Using the number honestly

A low cost per view does not make a clip good. If only 10 people watch a clip that took a week to script, the cost per viewer is high in time even when the generation fee is small. Track views and completions, and remove clips nobody watches. The table is a way to compare production choices, not a forecast of your audience.

Also keep the denominator real. Count unique viewers who play the clip, not page loads.

If you want a quick feel for scale, a clip that 500 people watch costs 3.68 / 500 = $0.00736 per viewer on standard. Even at max ($11.00) the figure is $0.022 per viewer at 500. Production time, not generation, usually dominates.