Synthesia launches Syren, a prompt-to-video agent that assembles avatars with motion graphics

Synthesia announced Syren on 9 October 2026 as a new AI video creation tool, launched in beta. According to the company, a Syren agent writes the video as code and Synthesia's in-house rendering engine plays it back live in the browser, with no installation required. Users can refine the output through chat-based editing.

Syren accepts a text prompt, a preset, or uploaded material including documents, images, videos, PowerPoint files and YouTube links. Output is up to three minutes long, horizontal or vertical, and can include motion graphics, animation, avatars, voiceover, music and b-roll. Synthesia says the feature is credit-based and that launch credits let every customer create a first video; the company's separate Syren update page states that every user receives 2,000 Syren credits, expiring after one month. Syren is available to self-serve free users and all Enterprise customers, reachable from a Syren button in the left-hand navigation, and Enterprise administrators can disable it at the organisation level.

Why it matters for digital humans: Syren marks a shift in how avatar vendors position the avatar itself. In Synthesia's scene-by-scene editor the avatar was the product; in Syren it is one layer that an agent assembles alongside graphics, b-roll and music, chosen by the agent rather than placed by the user. For practitioners that trades control for speed — useful for first drafts and internal explainers, harder to govern where brand, script and on-screen claims must be signed off. Note also that avatar and language counts on Synthesia's own launch page are inconsistent (240+ avatars in one section, 230+ in another), the credit economics are not published, and the whole feature is beta, so throughput and render quality claims are company-reported and untested at scale.