Synthesia built its entire business on the idea that a video is worth a thousand meetings. Its newest product argues the opposite: that watching a video barely teaches you anything at all. Roleplay Sessions drops the passive avatar and puts it across the table from you instead, arguing back during a sales pitch, pushing back on a performance review, then scoring how you handled it.
Here’s the thing. Synthesia has spent years selling companies on AI-generated training videos, cheap, fast, and scriptable. This week it quietly admitted that videos alone don’t change behavior. Its own launch materials point to learning research showing that most corporate training informs and demonstrates, but rarely gets anyone to actually do the thing under pressure. Roleplay is Synthesia’s answer to that gap, and it’s a genuinely useful shift if you manage a team that needs to practice hard conversations before having them for real.
What Synthesia Roleplay Sessions Actually Does
Launched this week, Roleplay Sessions is the first product under a new “Sessions” platform that Synthesia plans to expand into interviews and candidate screening. Employees log in and practice a specific scenario, a sales pitch, a difficult customer complaint, a tense performance review, with an AI avatar built to talk back, object, and stay in character. Once the conversation ends, the system scores the employee against a rubric and hands managers analytics across the whole team.
Synthesia CEO Victor Riparbelli put the reasoning simply: “Video performs way better than text or sending out a document.” But reading and watching only gets someone so far. Practicing the actual conversation, with feedback attached, is what changes how someone performs when it counts.
The avatars and voice tech are Synthesia’s own. The reasoning layer underneath, the part that argues back like a real customer or a skeptical manager, runs on OpenAI’s models. That combination is what lets Synthesia pitch this less as an avatar company and more as a performance-management tool with a video front end.
The Upside
- It shifts training from watching to doing. Reps and managers rehearse the actual conversation instead of absorbing another slide deck, which is the difference between knowing a script and being able to use it under pressure.
- Managers get real data, not gut feel. Instead of guessing who’s ready for a big client call, leaders can see scored performance across an entire sales team or leadership cohort in one dashboard.
- It works with what a company already has. Teams can build their own scenarios or hand Synthesia’s consultants their existing training documents and let them build a custom program around it.
- It’s built to scale beyond sales. Synthesia is already planning versions for job interviews and candidate screening, so this isn’t a one-trick pitch-practice tool.
Where It Falls Short
- It’s enterprise-only, for now. Small teams, solo founders, and individual learners are shut out until Synthesia opens this up to smaller businesses and prosumers, which it says is still months away.
- The reasoning engine isn’t Synthesia’s own. The conversational intelligence runs on OpenAI’s models, so how the AI “thinks” during a roleplay isn’t something Synthesia fully controls or can easily differentiate.
- Pricing isn’t public. You won’t find a self-serve checkout here. Everything runs through a sales conversation, which slows things down if you just want to try it out.
Who This Is Actually Built For
- Sales leaders who need every rep pitch-ready before a real call, not just script-familiar.
- L&D and enablement teams at large companies already running Synthesia’s video product who want to add practice, not just content.
- HR and people teams handling manager training for tough conversations like performance reviews and layoffs.
- Recruiting teams, once the interview and candidate-screening features Synthesia has planned actually ship.
- Not a fit for solo learners, freelancers, or small businesses wanting to dabble. Wait for the wider rollout.
Pricing
Synthesia hasn’t published pricing for Roleplay Sessions. It’s an enterprise product sold through a sales conversation, and current customers include some of the largest companies in Europe and the Fortune 100, so expect enterprise-scale quotes rather than a self-serve plan. For context, Synthesia’s core video platform runs free for light use, then scales up through paid tiers that add avatars, minutes, and API access, with custom Enterprise pricing at the top. Riparbelli has said the goal is to open Roleplay to small businesses and educational institutions as inference costs keep falling, so a lighter-weight plan may show up later this year.
Alternative Tools
If your team needs AI roleplay practice today and doesn’t have an enterprise budget or a Synthesia relationship already, a few tools cover similar ground:
- Second Nature AI focuses squarely on sales and service roleplay, with strong multilingual support and LMS integrations for teams that still run certifications.
- Yoodli is the more affordable, broader option, covering sales pitches, interviews, and public speaking, with a genuinely useful free tier.
- Hyperbound is built specifically for cold-call and outbound sales practice, trained on a large library of real B2B sales calls.
None of these give you Synthesia’s video-avatar polish or its plans to expand into interviews, but they’re accessible right now without an enterprise sales call.
The Bottom Line
Synthesia just bet that the next phase of enterprise AI isn’t about generating better content, it’s about proving the content actually changed how someone performs. If you’re managing a sales or leadership team and you’ve been feeding people slide decks and hoping it sticks, Roleplay Sessions is worth a look once it’s available at your company’s scale. If you need something today, Second Nature, Yoodli, or Hyperbound will get your team practicing without the enterprise sales cycle.
If you manage a team that keeps fumbling the same hard conversation, send them this. Their next tough call might go better because they already had it once, with an AI that pushed back first.
Sources and further reading

Leave a comment