Griffin-Lite is waitlist only: what can an avatar team ship now?

Short answer

You cannot ship Griffin-Lite this week. Tavus's own announcement says Griffin-Lite is a research preview for select trusted testers and that it "will not be available for use for customers at this time", with a waitlist form for access. If your plan needs a talking presenter in the next few days, the practical route is a rendered clip, not a live call. Sume Avatar 1.0 makes script-driven talking videos through a public API, and the docs describe the full flow.

What a waitlist means for a launch date

A waitlist with no published price or service terms cannot anchor a schedule. You do not know when access opens, what a minute costs, or what the safety conditions will be. The Tavus page says it is working on safe disclosure features before general availability, so even the first release may arrive with rules you have not seen.

• Do not put a Griffin-Lite date in a customer commitment.
• Do not budget a per-minute figure: the page gives none.
• Do keep the waitlist request open if a live, two-way avatar is a real requirement.

What you can ship now with Sume

Sume Avatar 1.0 is a two-step flow. You create a reusable avatar from a prompt, a profile, or a reference image, then generate a talking video from a script. Each request is a job you poll. This is a recorded clip, not a live conversation: the viewer cannot talk back. For announcements, onboarding, product explainers, and ad spots that is often enough.

Cost of a first week

Per-second rates on the Sume pricing page are $0.184 at standard, $0.245 at plus and $0.55 at max, and an avatar creation costs $0.95. A concrete first week: one avatar ($0.95) plus five 30-second clips at plus. Five clips at 30 seconds is 150 seconds, and 150 x $0.245 = $36.75. Add the avatar and the total is $37.70.

When to wait for live instead

Choose live only if the use case needs the avatar to listen and answer in real time. A rendered clip cannot do that. If you only need the same message delivered to many people, a rendered clip you can review before publishing is simpler and has a known price. Check the Tavus page again when it publishes terms, and decide then.

A middle path is to ship rendered clips now and keep the waitlist request in. If live access arrives later, the scripts, the approved presenter look and the disclosure wording you settled on still carry over, because they are decisions about your content rather than about a tool. The Tavus page also says developers can build today with its existing Phoenix, Raven and Sparrow models through its API or its PAL Maker platform, so a live product is not blocked on Griffin-Lite itself; that is a separate vendor decision outside this post.

A one-day plan

Morning: write three scripts of 20 to 30 seconds and pick one fixed disclosure sentence. Midday: create the avatar from a prompt, which costs $0.95, and generate one clip at standard to check pacing. Afternoon: render the approved scripts at plus. Because each request is a job, you can poll them and keep working, and the finished videos arrive as public media URLs. Nothing in this plan depends on a vendor waitlist.