Synthesia’s new prompt-to-video agent writes each video as code and renders it live in the browser. The company is positioning that engine — reported as “Sirén” — toward interactive AI video, beyond fixed avatar clips.
2026-10-26 · by Moe Ameen
On October 9, 2026, Synthesia launched Syren, a new AI video agent, in beta. You describe a video in a chat prompt — or start from a preset or existing material such as a document, deck, image, video, or YouTube link — and Syren plans and builds it: script, motion graphics, animation, avatars, voiceover, music, and b-roll. The headline is “On-brand video from one prompt,” and the company says Syren renders a finished video in minutes, up to three minutes long, in horizontal or vertical format.
What is technically notable is how it renders. Rather than assembling scenes on a timeline, Syren’s agent writes the video as code, and an in-house rendering engine plays that code back live in the browser as the video is built — nothing to install. You make changes by chatting with the agent (“swap the avatar,” “restyle that scene”) instead of editing by hand. Synthesia says the same engine has powered motion graphics inside its product since June 2026. Coverage of the launch refers to the engine by the name Sirén; Synthesia’s own pages call it an in-house rendering engine rather than naming it, so treat the name as reported, not official.
The strategic shift is the framing. Synthesia built its business on avatar training video made in its Studio editor, where every line and scene is controlled for compliance and onboarding. Syren is the opposite lane — fast, designed, creative output for product launches, explainers, how-tos, marketing, and ads — and the company is positioning the code-native, live-rendering approach as the foundation for interactive AI video, where a clip could branch and respond rather than play back as a fixed file. For now Syren outputs a linear video; interactive is the direction, not a shipped Syren feature. Synthesia’s interactive products today are separate: Interactive Avatars, Roleplay Sessions, and Survey Sessions.
Syren is in beta for self-serve, free, and Enterprise customers, reached from a Syren button in the left-hand navigation; Enterprise admins can turn it off organization-wide. It runs on credits — launch credits cover a first video, and Synthesia says a two-minute video with a couple of rounds of edits typically uses on the order of 1,000 to 2,000 credits — and free-plan videos carry a watermark. Brand control is still early: Syren applies your colors now, with full brand kits (typography, animation styles) and learning a brand from your existing videos flagged as coming. Personal avatars and cloned voices still require Synthesia Studio. Treat the specifics as a beta snapshot and check the live page.
The useful move the day Syren ships is not to wait for the interactive future — it is to solve the gap Syren leaves open the minute it hands you an MP4: distribution. Syren renders one polished, on-brand video up to three minutes long and stops. [Kompozy](/) is the layer that turns that clip into a week of posts. Drop a Syren explainer or ad into Kompozy’s [Clipped Shorts](/glossary/clipped-short) and it cuts the long render into vertical, platform-native cuts with word-synced captions that carry on the feeds where Syren’s generated voiceover plays silent. From the same idea, Kompozy generates the formats Syren does not — a [HyperFrames](/glossary/hyperframes) carousel, quote graphics, per-platform text, a blog, and a newsletter — all held to one [Persona Brief](/glossary/persona-brief) voice.
Then [Autopilot](/glossary/autopilot) schedules and publishes the set across the eight social platforms plus blog and email, behind a per-post review. Kompozy also generates its own avatar video — [Persona Shorts](/glossary/persona-shorts) and HeyGen-based formats — so you are not locked to one vendor’s render for the talking-head half. The pattern holds whatever Synthesia ships next: Syren (and whatever interactive layer the Sirén engine grows into) makes the on-brand video; Kompozy clips it, multiplies it into every other format, and gets it live everywhere. To weigh the avatar-video tools themselves, see our [Synthesia](/ai-tools/synthesia) and [HeyGen](/ai-tools/heygen) pages.
Syren is Synthesia’s AI video agent, launched in beta in October 2026. You describe a video in a prompt (or start from a preset or uploaded context), and an agent writes it as code and renders it live in the browser — building motion graphics, animation, avatars, voiceover, music, and b-roll into an on-brand clip up to three minutes long. You refine it by chatting with the agent rather than editing scenes by hand.
They are related but not identical. “Syren” is the product — the AI video agent you use. “Sirén” is the name reporting has given the in-house rendering engine underneath it, which plays the agent’s code back as live video. Synthesia’s own pages describe an in-house rendering engine without naming it, so treat Sirén as the reported engine name behind the Syren product.
Not today. Syren outputs a linear video that plays back as a fixed file. Synthesia is positioning the code-native, live-rendering engine as the foundation for interactive AI video — clips that could branch and respond — but that is the company’s direction, not a current Syren feature. Synthesia’s interactive products in 2026 are separate: Interactive Avatars, Roleplay Sessions, and Survey Sessions.
Syren is in beta and runs on credits. It is free to try for self-serve users with launch credits for a first video (free-plan output carries a watermark), and it is available to Enterprise customers whose admins can disable it organization-wide. Synthesia says a two-minute video with a couple of rounds of edits typically uses roughly 1,000 to 2,000 credits. Verify current pricing on synthesia.io.