Synthesia Roleplay Sessions vs devlin.ai: 2026 Comparison
By Devlin Peck · Updated
Part of the AI in Instructional Design guide
Synthesia Roleplay Sessions and devlin.ai both let employees practice high-stakes conversations with an AI character that talks back, pushes back, and scores the attempt. The choice comes down to where practice should live. If your organization already runs on Synthesia and you want avatar video practice inside that platform, Roleplay Sessions is the natural pick. If you want voice-first practice embedded in the courses and LMS you already have, with no learner accounts to manage, that is what devlin.ai is built for.
Full disclosure before anything else: devlin.ai is my product. I will tell you where each tool fits, including where mine falls short, and I will keep the comparison to what both companies have published. Claims about both products were checked against the vendors' live published materials as of August 2026; Synthesia's docs are changing quickly during this rollout, so treat their side as a dated snapshot. This piece is part of my full guide to AI in instructional design.
What is Synthesia Roleplay Sessions?
Roleplay Sessions is Synthesia's AI practice product, launched in July 2026, where employees rehearse conversations like sales pitches, performance reviews, and customer complaints with one of Synthesia's Interactive Avatars. The avatar improvises in character, then the session is scored against a rubric, and results feed manager dashboards with a team-wide view, transcript drill-downs, and session replays. At launch it supports practice in English, German, Spanish, and French, with more languages planned, per the launch announcement.
It is also the first release on Synthesia's new Sessions platform, which TechCrunch reports will expand into job interviews and candidate screening. This is a company repositioning from making videos toward measuring performance, and it tells you the product will get sustained investment.
The build workflow is deeper than the launch press suggests, and it is worth knowing before a demo. Per Synthesia's Sessions documentation, designers can create a roleplay three ways: through an Assistant that takes supporting files and URLs as context, from templates, or by adding a Roleplay scene (plus a Feedback scene) to a video. Personas draw on a library of six base personalities and can be given objections to raise mid-conversation. Scoring runs on skills with adjustable weights and up to six scoring levels each, plus configurable pass scores and retries. Delivery has four options: Synthesia's Learner Portal, a share link, an HTML embed, or SCORM export to your LMS. One detail from their docs to note for reporting: SCORM delivery and the portal's own tracking do not sync with each other, even for the same learner.
Synthesia's launch materials also report early results: roughly a 30 percent average skill uplift at a global recruiting firm and about 70 percent of a Fortune 10 sales team coming back for voluntary practice. Those are company-reported launch figures, not independent studies, but they say something real about how learners respond to conversation practice.
On availability: as of August 2026, Roleplay Sessions is sold through enterprise licensing, pricing has not been published, and existing Synthesia customers are getting access first, with broader availability to smaller businesses and individuals planned, per TechCrunch's coverage. That matches what I saw at the August 2026 launch webinar: attendees (myself included) could open the demo roleplays, but many could not yet find a create flow in their own accounts, and licensing and pricing questions were directed to account teams. None of this is unusual for a staged enterprise rollout. It does mean that if you are evaluating it this quarter, your first question is what is actually enabled on your plan today.
What is devlin.ai? (full disclosure: this one is mine)
devlin.ai is a simulation builder: describe a scenario in plain English and you get a working simulation you can deliver wherever your learners already are. Instructional designers define the character, the situation, and the evaluation criteria, then embed the result in Articulate Storyline or Rise, drop it into any LMS via SCORM or xAPI, or share it as a hosted link. Learners talk with the character by voice or text in any of 16 supported languages, get scored against the rubric the designer defined by an evaluation engine we regression-test so its accuracy stays reliable, and the results flow back into the course, including pass/fail and score variables inside Storyline itself. Scenarios can be grounded in your own documents, so a character quotes your return policy rather than a plausible-sounding invention.
On the admin side, there is a full analytics dashboard with exports and Power BI sync, SSO/SAML, and full workspace export, so your work is never locked in. Learners don't create accounts by default (although on enterprise plans we can do learner SSO); they open the course or the link and start talking. Getting started is self-serve, with no sales call required, and more than 1,700 instructional designers are building on the platform as of August 2026. The fastest way to judge any of this is to make a simulation yourself: describe a scenario your team actually trains on, and try what comes back.
Where it falls short: there is no avatar video. That is a deliberate tradeoff (voice keeps the learner's attention on the conversation or other visuals rather than a rendered face), but if your stakeholders want a person on screen, Synthesia is ahead there and likely will stay ahead. devlin.ai launched in 2026, so it is a young product. It is a practice layer, not a video platform, so you will keep making your training video somewhere else. And as of mid-2026, the simulations are optimized for character-based conversation practice (an upset customer, a coaching conversation with a defensive direct report), not vocational or safety walk-throughs.
On the trust side, learner voice audio is not retained; session review works from transcripts.
How do the two compare side by side?
The structural difference is that Synthesia is a destination platform that now includes practice, while devlin.ai is a practice layer that lives inside the learning experiences you already ship. This table is a breakdown of the two vendors' own published materials, checked in August 2026.
| Synthesia Roleplay Sessions | devlin.ai | |
|---|---|---|
| Practice format | Video conversation with an Interactive Avatar | Voice or text conversation, no avatar |
| Where learners practice | In Synthesia's Learner Portal, by share link or HTML embed, or via SCORM export to your LMS, per their docs | Inside your courses (Storyline, Rise, LMS) or a hosted link |
| Learner access | Learners sign in with their Synthesia credentials per their materials; confirm what a seat means under your license | No learner accounts; learners open the course or link |
| Scoring | Skills with adjustable weights and up to six scoring levels, pass scores, session replay, transcripts, and team dashboards | Designer-defined rubrics with transcripts (voice audio is not retained); results write to SCORM/xAPI and Storyline variables |
| Feedback experience | A Feedback scene where a conversational avatar coach reviews the attempt and learners can ask follow-up questions | AI coaching plus a written evaluation against each criterion, with the full transcript |
| Retries | Designer-configurable per session | Attempt data tracked in the dashboard; test attempts can be deleted |
| Grounding in your content | Assistant creation takes supporting files and URLs as context | Scenario setup plus knowledge base grounding in your uploaded documents |
| Platform context | One product in an AI video platform | The whole product; video stays with your video tool |
| Languages | English, German, Spanish, French at launch, more planned | 16 languages |
| Pricing model | Enterprise licensing; talk to your account team | Usage-based AI credits on Team and Enterprise plans |
| Getting started | Enterprise rollout; access through your Synthesia account team for now | Self-serve; sign up and build the same day |
On the scoring row, here is how devlin.ai's side works under the hood, since I built it: a master prompt takes in the criteria you define, each with a point value you set, and evaluates the full conversation transcript against them, assigning full credit, partial credit at 50 percent, or zero credit per criterion. Synthesia's weighted-skills model in their docs is a different route to the same goal. Either way, insist on seeing the rubric authoring screen in any demo, because the rubric is where your standards live.
When should you choose Synthesia Roleplay Sessions?
Choose Roleplay Sessions if most of these are true:
- Your organization already licenses Synthesia and your learners already work inside it.
- Avatar presence matters to your stakeholders, and a voice-only experience would be a hard sell.
- You want practice as a managed destination, with manager dashboards as the primary reporting surface.
- Your learners practice in English, German, Spanish, or French.
- You would rather consolidate with one large vendor than add a specialist.
That last point may make a lot of sense for your team. Vendor consolidation is a legitimate reason to pick the bigger platform, and I would rather say so than pretend otherwise.
When should you choose devlin.ai?
Choose devlin.ai if most of these are true:
- Practice needs to happen inside the courses you already have or are building, not on another platform.
- You are rolling out to hundreds or thousands of learners and do not want to manage accounts or per-learner platform access for any of them.
- Your scenarios must stay grounded in your actual policies, SOPs, or product documentation.
- You are training phone-heavy roles (contact centers, sales calls, telehealth) where voice practice without a face on screen is the realistic condition.
- You want to pilot this quarter: sign up is self-serve, and there is no sales cycle required to start building.
If you are earlier in the decision than this, before vendor selection, start with when AI role-play works and when it does not.
What does a founder-led vendor actually change for you?
It changes who is in the room. Buy from a large platform and you get procurement comfort, an account team, and a product suite; buy from a founder-led company and you get the person who built the product sitting in your implementation conversations. Neither is automatically better, so weigh both sides.
What Synthesia's size buys you is real: mature enterprise processes, a large ecosystem around an established platform, and the ability to consolidate video and practice under one contract with one vendor to manage.
What founder-led buys you is speed and depth. Requests do not queue behind a product committee. Roadmap conversations happen with the person who ships the code, and solutions get tied to your business goals rather than a standard deployment playbook. This is what the buying experience is like when the builder is on the call: requests get scoped and shipped directly, without a product committee in between. We also publish everything we ship, with dates, on the devlin.ai site, so you can check the cadence yourself rather than take my word for it. In the first week of August 2026 alone, three platform updates went out.
The tradeoff cuts the other way too: a smaller vendor asks you to bet on fewer people. That is exactly why the security question in the checklist below matters, and why devlin.ai includes full workspace export. If you ever leave, your work leaves with you.
And the reason I am the one answering roadmap questions is simple. I built devlin.ai because what excites me most about AI is not recreating old formats faster. It is letting a learner practice a difficult conversation with a character who has a personality, emotional responses, and motives, in a conversation that adapts in real time instead of a branching scenario with three pre-written paths.
Can you use both?
Yes, and for Synthesia-heavy organizations this is often the right answer rather than a compromise. Use Synthesia for what it is best at, producing the instructional video your program needs, and embed devlin.ai simulations inside the same courses for the practice and assessment layer. The two do not compete for the same slot in that architecture.
One more thing worth saying about the category: Synthesia entering practice is good news for everyone who has been arguing that practice-based training beats content delivery. It matches what I hear directly. In my conversations with hiring managers, learning leaders, directors, and senior IDs over the past several months, nearly everyone is trying to add AI-powered scenarios and role-play to their learning experiences, and most have not figured out how to do it yet. When a large platform pivots from making videos to measuring conversations, it confirms the direction. The competition is on depth from here.
What should you ask either vendor before you buy?
Send both vendors the same five questions, verbatim, and compare the answers in writing:
- Do learners need accounts or licenses, and what does a rollout to 1,000 learners cost in total?
- Can practice run inside our LMS and authoring tools, and how do scores get back to us (SCORM, xAPI, or API)?
- Can we define our own scoring rubric, and can reviewers see full transcripts?
- What security attestations can you share today (for example a current SOC 2 report), and how long is conversation data retained?
- What happens when a learner goes off script or tries to break the character?
A longer version of this checklist, including the build-vs-buy math, is in how to evaluate AI simulation tools.
What are the alternatives to Synthesia Roleplay Sessions?
Beyond devlin.ai, two dedicated practice platforms come up most often: Second Nature, which focuses on AI role-play for sales and revenue teams, and Yoodli, which started in speech coaching and has expanded into broader communication roleplay. I have not run either through a full hands-on test, so I will not rank them; the platform-versus-layer question above is still the right first filter for both. My wider tool rundown is in the AI tools I actually use.
The non-AI alternative is the branching scenario you build yourself in a tool like Storyline. It is still the right call for compliance-critical paths where every word must be authored, and the tradeoffs are covered in my AI role-play briefing linked above.
Frequently asked questions
Do learners need an account to practice?
With devlin.ai, no. Learners open the course or the link and start talking; there is nothing to sign up for. With Synthesia Roleplay Sessions, learners sign in with their Synthesia credentials per the company's materials, so confirm with your account team what a learner seat means under your license.
Can AI roleplay run inside Articulate Storyline or an LMS?
devlin.ai is built for this: simulations embed in Storyline and Rise, package for any SCORM or xAPI compliant LMS, and write results back to Storyline variables. In my own builds, the feedback slide pulls the sim feedback, score, and pass/fail variables straight onto the slide, and the next button stays gated until the evaluation-complete variable flips to true, so a roleplay and a knowledge check can post as one scored event. Synthesia's materials describe SCORM export to your LMS as well, and their documentation notes that SCORM delivery and their portal's tracking do not sync with each other; the question to put to their team is whether the roleplay can live inside one of your existing courses and contribute to that course's single score, or ships as its own standalone object.
What languages does Synthesia Roleplay Sessions support?
English, German, Spanish, and French at the July 2026 launch, with more planned, per Synthesia's launch materials. Re-check their documentation for the current list before you commit a non-English program.
Can you buy Synthesia Roleplay Sessions today?
As of August 2026, Roleplay Sessions is sold through enterprise sales, pricing has not been publicly disclosed, and existing Synthesia customers are getting access first, with broader availability to smaller businesses and individuals planned, per TechCrunch's launch coverage. Check Synthesia's documentation and your account team for the current status on your plan.
How much does AI conversation practice cost?
Synthesia Roleplay Sessions is sold through enterprise licensing, so pricing comes from their account team. devlin.ai prices on usage-based AI credits within its Team and Enterprise plans, so cost tracks how much your learners actually practice. To make this modelable rather than vague: a voice conversation runs roughly 130 to 150 credits per minute, and a fully completed 10-turn text simulation usually lands around 70 credits. Voice is what makes the practice feel real, so build your program around it; these numbers exist so you can forecast spend for your learner counts before you commit. For the category-level cost math, see the cost section of my AI role-play briefing.
How is this different from just using ChatGPT?
A general assistant can improvise a character, but neither of these products is competing with that. What you are buying is the layer around the conversation: designer-defined rubrics, consistent scoring, transcripts for reviewers, retry and pass logic, and results reporting into your LMS. Synthesia's own feature page draws the same line, and it is the right one: without measurement and reporting, roleplay is an activity, not training.
The Bottom Line
If your learners already live in Synthesia and your stakeholders want avatar video, pick Roleplay Sessions and pressure-test the rollout details on your plan. If practice needs to live inside the courses and LMS you already ship, at learner counts where per-learner accounts stop making sense, pick devlin.ai. If you run a large Synthesia video program, the strongest architecture is often both: their video, your courses, and the devlin.ai practice layer inside them.
Whichever way you lean, send both vendors the five questions above before anyone talks pricing. The answers will make the decision for you faster than any demo. And if you are still earlier in the journey, start with the full guide to rolling out AI simulations.