Company launches developer APIs, releasing state-of-the-art Phoenix model that powers video generation from text using your digital replica

Tavus, a leading generative AI video research company, today announced the beta launch of the Tavus Developer Platform. Now the company’s proprietary models will be accessible via APIs to product teams looking to leverage AI to build novel engagement experiences across sales, marketing, and online learning. The first available Replica API utilizes Tavus’ breakthrough Phoenix model, which powers digital replica and text-to-video generation capabilities; it gives end-users the ability to rapidly create authentic videos of themselves on any topic, with only a script and no recording or studio production.

We’re thrilled to open the doors to our cutting-edge models, empowering developers to craft the next wave of captivating video experiences. —Hassaan Raza, Co-founder and CEO of Tavus

The release of the developer platform comes after Tavus raised $18 million in Series A funding late last year led by Scale Venture Partners, with Rory O’Driscoll joining the company’s board of directors. Sequoia Capital, Y Combinator, HubSpot, and others also participated in the round. Funds will be used to continue accelerating product and model development. Tavus also plans to expand sales in enterprise and mid-market with demand for its SaaS application coming from a growing range of use cases in banking, real estate, automotive, and health care.

“Companies need to meet customers where they are, and today’s customers are consuming video. To make it engaging, it has to feel personal, and to make it scalable, it needs to be automated,” said Rory Driscoll, Partner at Scale Venture Partners. “The team at Tavus, led by co-founders with a strong vision, has made this possible and we’re excited to partner with them as they put the technology in developers’ hands to create personalized video experiences with generative AI.”

Marketing Technology News: MarTech Interview with Allison Breeding, CMO at Apptio, an IBM company

Tavus pioneered generative AI video in 2020, initially leading the industry breakthroughs in large-scale personalized video campaigns using lip syncing and audio blending techniques. Today, the Tavus team moves its technology well beyond existing capabilities, releasing its advanced Phoenix model. Phoenix generates exceptionally realistic talking head videos, complete with natural face movements and expressions accurately synchronized with input. Developed by Tavus, this novel approach bypasses traditional methods and constructs dynamic, three-dimensional facial scenes using neural radiance fields (NeRFs). The result is unprecedented realism and higher fidelity videos compared with open-source models.

“As we’ve been refining our product and advancing our models, we’ve received requests for API access from a broad spectrum of developers, from emerging startups to industry giants,” said Hassaan Raza, Co-founder and CEO of Tavus. “We’ve always been committed to quality research, and we see limitless potential for our technology’s application. We’re thrilled to open the doors to our cutting-edge models, empowering developers to craft the next wave of captivating video experiences.”

Marketing Technology News: What Can Help Differentiate Your B2B Marketing Case Study?

Tavus is the only platform that offers developers end-to-end models and APIs of this quality for generative talking-head videos. The first API to be released is its Replica API in Beta. Available now, this API enables photo-realistic digital replicas and text-to-video generation.

Several more APIs are coming soon, including:

• Lip Syncing API: In-place or full lip syncing.
• Dubbing API: Translation and dubbing with seamless lip syncing.
• Video Campaign API: Video orchestration for scaled personalized video generation.

Tavus is built with privacy and security always front-of-mind. Tavus’ technology cannot be used to replicate people without their explicit permission. It employs a suite of safety checks including voice identification and matching consent with training data to ensure that only you can create a replica of yourself and hold the keys to your likeness.