Synthesia Sessions launches live interactive avatars
Synthesia launched Sessions, turning AI avatars into real-time video participants. Learn how it works, what it costs, and what questions remain open.
The platform replaces pre-rendered video clips with live two-way calls for surveys and corporate training.
On October 1, 2026, Synthesia rolled out Synthesia Sessions, an enterprise platform designed to put artificial intelligence avatars directly into real-time video calls. The launch was signed by the company’s chief technology officer, Peter Hill. The release targets a common operational bottleneck: running guided conversations at scale without keeping human staff on call around the clock.
For corporate teams following video generation tools, this marks a practical pivot. Until now, the British startup built its core business on exporting finished video files. With Sessions, synthetic presenters can now hear incoming speech and reply during a live call.
From pre-rendered clips to live calls
Before Sessions, creating synthetic video avatars followed a linear path. A user typed a script into a text box, chose a digital face, and waited for cloud servers to compile an MP4 file. That setup worked well for corporate handbooks and compliance explainers, but it remained entirely one-directional. If a worker had a question halfway through, the clip could not answer.
With Sessions, Synthesia moves the format closer to a regular web meeting. The system captures the user’s microphone audio, processes what was said, and streams back video responses with matching lip movements in fractions of a second.
What the platform delivers today
The platform launched with two operational features available for corporate accounts.
The first is Survey Sessions, designed to replace static web questionnaires. Instead of clicking through radio buttons on a screen, respondents talk to an avatar that asks follow-up questions tailored to their earlier comments. According to the company, the tool generates summary dashboards with transcripts, sentiment scores, and recurring topics once the call wraps up. Every plan includes an entry allowance of 3 free responses per month.
The second feature is Roleplay Sessions, aimed at sales practice and support coaching. Workers interact with a synthetic persona configured to simulate friction, such as handling unhappy customers or fielding difficult sales objections. Synthesia piloted an early iteration of this coaching tool in July 2026, as reported by TechCrunch, and has now brought it into Sessions with broader language support. The expanded roster covers Portuguese (both Brazilian and European), Italian, Polish, Chinese, Arabic, Hindi, Dutch, Swedish, and Turkish. Each account receives 10 free roleplay sessions per month.
How Synthesia Sessions works in practice
To keep conversations from stalling on awkward pauses, Synthesia tuned low-latency visual models focused on lip synchronisation. The company also launched an Interactive Avatar API. This developer interface allows engineering teams to connect Synthesia’s visual avatar rendering to their own large language models and speech synthesis systems.
In practice, that means the on-screen persona handles the visual delivery, while the underlying reasoning engine deciding what to say can run entirely inside an organization’s own technical stack.
Limits, costs, and undisclosed details
Behind the product launch, several key operational details remain unaddressed. The main question involves the underlying language model: Synthesia’s October 1 announcement did not reveal which foundation model powers conversational reasoning inside turnkey sessions. While TechCrunch previously reported the startup relied on OpenAI technology in earlier builds, Synthesia did not confirm its current supplier in the launch materials.
Free monthly quotas are also narrow. Three survey responses and ten roleplay runs are only enough for small-scale testing. Any organization planning regular production rollouts will have to negotiate custom rates with the company’s sales reps, as extra volume pricing is not listed publicly.
Furthermore, a third promised module called Expert Sessions — designed to handle internal documentation and employee troubleshooting — remains listed as in development, with no release schedule or pricing details attached.
What this means for teams
Synthesia Sessions offers a sensible upgrade for organizations already using video avatars that need basic feedback loops in surveys or rookie sales practice. For now, it fits companies looking to pilot spoken interviews or give new hires a safe space to rehearse client calls without tying up senior staff. For higher-stakes workflows, the lack of published usage pricing and unanswered questions about the underlying language model call for measured testing before signing enterprise commitments.
Sources
Official announcements
Independent reporting