Synthesia unveils 7 new features at once in July, boosting roleplay and dubbing

Synthesia is a company that uses AI to create videos of people appearing to speak on camera. Founded in London in 2017, it has focused on "talking head" videos — the kind used for corporate training or internal announcements — that can be made without cameras, microphones, or film crews. It's often grouped alongside general-purpose video generation models like Sora or Veo, but what it produces is different. Its avatars aren't imagined faces either; they're based on filming real actors with their consent.

Seven updates rolled out in July

The video walks through the features released over the course of July in chronological order. Summarized in table form:

Feature	Key details
Roleplay Sessions	Conversation practice with real-time coaching and feedback
Dubbing 2.0	130 languages, lip-sync matched to native speaker mouth movements
Customized Avatars	Reflects brand identity such as uniforms and colors
Voice Variants	Overall more natural-sounding voices
Motion Graphics	Edit text, titles, and stats directly within a project, with built-in translation
Auto Translation	Automatically applies translation to motion graphics elements
AI Video Presets	Generates polished B-roll in seconds without prompts

Synthesia said Dubbing 2.0 expands language support to 130 languages while improving lip-sync to match the mouth movements of native speakers. Customized Avatars lets companies apply their own brand elements — uniforms, colors, logos — directly to avatars. AI Video Presets were described as letting users pick from pre-built styles to generate B-roll, rather than having to figure out what prompt to write.

How to try it

The entry point shown in the video is the project screen within the Synthesia platform. Specific menu steps weren't disclosed, but based on how each feature works, the likely flow is as follows:

1. Open a Synthesia project and choose a new video type — Roleplay Sessions for conversation practice, or Dubbing 2.0 if a multilingual version is needed.
2. In Roleplay Sessions, set up the conversation scenario you want to practice; the AI takes on the role of the other party and displays real-time coaching prompts during the exchange.
3. Send the finished video to Dubbing 2.0 to produce a lip-synced dub in any of 130 languages, matched to native speaker mouth movements.
4. For branded videos, use the Customized Avatars menu to create an avatar with adjusted uniforms, colors, and other elements.
5. When B-roll is needed, select a pre-built style in AI Video Presets to generate insertable footage within seconds.

The primary audience is teams producing corporate training or internal announcement videos. For example, when building an onboarding video for new hires, a team could use Roleplay Sessions to practice handling difficult customer scenarios, while simultaneously using Dubbing 2.0 to produce multilingual versions for overseas offices. Marketing teams could use Customized Avatars to create a greeter video featuring company uniforms, then fill in background B-roll with AI Video Presets to polish the final product.

Editor's take

What stands out in this batch of updates is that Synthesia is quietly expanding from being "a company that makes videos with avatars" into "a company that trains people to have conversations." Roleplay Sessions isn't just playback of a video — it's an interactive feature that delivers real-time coaching feedback. Until now, Synthesia's avatars were confined to one-way videos reciting fixed scripts; this feature marks a step toward avatars that respond to a user's input. In the corporate training market, roleplay has long been the job of human instructors or actors. This reads as a signal that AI avatars could take over that role.

Putting avatar platforms like this into actual practice tends to lead to the same conclusion every time: you can skip the camera crew, voice actors, and translation team, but the "naturalness" of the output still needs human review. Even with Dubbing 2.0 supporting lip-sync across 130 languages, real-world corporate training use often still requires a human check for culturally awkward phrasing or accent mismatches. This update is a clear step forward in language coverage and lip-sync precision, but it doesn't translate directly into full automation.

For domestic companies, it makes sense to first apply avatar platforms like Synthesia to internal training or onboarding videos. For brand advertising or external PR content, where subtle nuance matters more, it's safer to assume human involvement is still needed. Conversational features like Roleplay Sessions are worth testing first in areas requiring repetitive practice, such as training new hires to handle customer interactions or rehearsing sales scripts.

In the coming weeks, competing avatar platforms such as HeyGen and Hedra are likely to follow with similar real-time conversation and multilingual dubbing features. The next front in the avatar video market's competition is shifting from "how photorealistic" to "how naturally it can converse."