Synthesia releases a ChatGPT plugin that generates avatar video inside the chat

Synthesia announced on 3 November a plugin for ChatGPT that generates video without leaving the chat. A user describes the video they want, or shares a document, deck or spreadsheet, and the plugin returns a finished video that can then be revised in the same conversation. The default output uses Syren, Synthesia's prompt-to-video format, which combines motion graphics, avatars, voiceover, music and b-roll; asking for it in Studio instead produces an avatar delivering the script word for word, with scenes editable afterwards in Synthesia's video editor. The company positions that second path at training, onboarding and compliance material. The plugin can also search a user's existing video library and surface video details, and it can be chained with other connected ChatGPT plugins to run once, on a schedule, or on a trigger such as an incoming Slack message or email. Synthesia says the plugin is live on every plan and cites 240+ avatars and 140+ languages. It does not state usage quotas, maximum video length, pricing tiers or any ChatGPT plan requirement.

Why it matters for digital humans: The notable part here is distribution, not the model. Avatar video has mostly been produced in a destination app that someone has to open deliberately; putting it behind a plugin makes it a step inside whatever assistant a team already has open, and the scheduling and trigger options make it an automated output rather than a considered one. That cuts both ways. It shortens the path from source document to published explainer, and it also removes the pause where a human read the script before an avatar wearing a company face said it out loud. Teams running avatar presenters under brand or compliance review should assume prompt-to-video pipelines will route around that review unless they design for it. The avatar and language counts are company-reported and inconsistent on Synthesia's own page, which cites 230+ avatars elsewhere, so treat them as marketing figures rather than a spec.