Can you buy Tavus Griffin-Lite? Presenter video options today

No. Tavus says Griffin-Lite is not available to customers at this time and is open only to select trusted testers as a research preview. If you need a presenter video this week, the realistic route is a scripted talking video, not a live conversation model. Sume Avatar 1.0 ships that: you create a reusable avatar, send a script, and get back an MP4 from a job.

What Tavus says about Griffin-Lite

The vendor page (Tavus: Griffin, read 2026-10-08) describes a Human Interaction Model announced on October 1, 2026. The facts that matter for a buyer are availability and the safety stance, not the headline study number.

What a scripted presenter is, and what it is not

Griffin is described as a real-time model: it reacts, interrupts and back-channels in a video call. A scripted avatar video is a different product. You write the words in advance, the job renders them, and nobody on screen can answer a question. For ads, explainers, onboarding clips and product walkthroughs that is usually what is wanted, because the output is a file you can review before it ships.

Sume does not list a real-time conversational video model. Its avatar surface is job based: submit, poll, then read the result URL. Read Models for the full family list.

What you can do on Sume today

The documented Avatar 1.0 flow is two steps. First create an avatar from a prompt, a structured profile or a reference photo. Then call the talking-video route with the avatar handle and a script.

• Create: POST /v1/avatar-1.0/generate with a top-level avatar_handle and an input of type prompt, props or photo.
• Render: POST /v1/avatar-1.0/talking-video with a script or ordered video_inputs, not both.
• Duration: Sume accepts scripts whose estimated length is 4-60 seconds inclusive.
• Quality: standard, plus (the default) or max.
• Preview: first-frame stills can be approved before the full render.

A sensible rule while the preview is closed

Plan around what you can run now. If a live, conversational presenter is part of your roadmap, treat Griffin-Lite as something to watch, and note that its own maker is still working on disclosure features. If you need a spokesperson clip, build it as a scripted job, keep the script reviewable, and label AI presenters where your channel or law requires it.

The details of each Sume step are on Create new avatar and Generate avatar video. Nothing on those pages depends on Tavus, and nothing here implies a partnership or that Sume resells Griffin.

How to decide this week

Ask three questions. Does the viewer need to talk back? If yes, a scripted video is the wrong tool, and you will have to wait for a live product or build the conversation yourself. Do you need to review every word before it reaches an audience? If yes, scripted wins. Is the clip short, under a minute? Then the 4-60 second Avatar 1.0 window is no obstacle.

Treat any vendor study as the vendor's own claim. Tavus reports its 48 percent figure from a study it ran with 54 participants, and the same page says wider release waits on safety work. That is useful context for planning, not a basis for a purchase today. Re-read the vendor page before you cite it again, since access terms can change without notice.

If you want a test of the scripted route, create one avatar from a prompt, render a 15-second script at the default plus quality, and judge it against the live-call experience you were hoping for. That small trial costs far less than a re-platform and tells you quickly whether a file-based presenter meets the need.