Can I use Tavus Griffin-Lite yet? Preview status and what to ship

No, you cannot use Tavus Griffin-Lite as a customer yet. Tavus describes it as a research preview "for select trusted testers" and states that it is "not available for use for customers at this time". The page offers an access request form, and shows a release date of October 1, 2026 (read 2026-10-05). If you need a talking avatar in a product this month, plan around a model you can call today.

Sume Avatar 1.0 is such a model. It makes a talking video from an avatar and a script through an API, with jobs, webhooks and a published price. That is a different product from a live call: it renders a clip, it does not hold a conversation. Tavus: Griffin is the source for every Tavus fact in this post.

What the Tavus page says

Here is what Tavus says about the preview, in its own words, next to the plain consequence for a builder.

Do not read the date as a launch

The last row matters. A date on a research page is easy to read as a launch. Tavus uses it for the preview, and the same page says the model is not for customers. Check the page again before you rely on any change, because this reading is dated 2026-10-05.

Keep a note of the date you last checked, since a page like this can change. This reading is dated 2026-10-05, and the Tavus page itself shows October 1, 2026 for the preview. If your plan needs a customer-ready live avatar, put a recheck on the calendar and design the rest of the product so that the live piece can be added later without a rewrite.

Which product for which job

Decide by use case, since the two kinds of product do different jobs. A live, two-way avatar belongs in a support desk, a tutoring session or a sales call where the person speaks and the avatar answers. A rendered clip belongs in an explainer, an ad, a product update or a welcome message where the words are known before the video exists.

• Live conversation, the viewer talks back: a full-duplex model such as Griffin-Lite, when it is open to you.
• A fixed message, the script is known: a rendered avatar video from Sume Avatar 1.0.
• Many versions of one message, per customer or language: render clips from a script template.
• Disclosure matters: add a plain on-screen line to a rendered clip, which you control today.

What to build on this week

On Sume the path from a photo to a video is two calls. Create an avatar from a photo with POST /v1/avatar-1.0/generate, then render with POST /v1/avatar-1.0/talking-video, with a script of an estimated 4 to 60 seconds. Quality is standard, plus (the default) or max, and output is 720p. See Create new avatar and Generate avatar video, both read 2026-10-05.

The Tavus page notes it is working on disclosure features; with a rendered clip you decide the disclosure text yourself. See the Griffin API post for a shorter view of the same question.

How to revisit this

When Tavus opens Griffin-Lite to customers, test it on the job it is built for. Run it side by side with a rendered clip on the same script, and decide whether the extra liveness is worth the added complexity and cost for your case. Until then, the safest plan is to ship something that works today and keep the live piece as a later upgrade. For the difference between the two designs, read real-time vs video avatar APIs.