Rapport vs Second Nature: Which AI Roleplay Platform Fits Your Team in 2026?

• Both platforms put a face in front of the learner, and the faces are made differently. Rapport animates a 3D avatar in real time from the learner's audio, at roughly 50ms. Second Nature ships video and moving avatars that arrived recently, and its own reviewers say they are not finished yet.

• Second Nature is further along on enterprise plumbing. SCORM and LTI with score passback, SCIM provisioning, SAML SSO, and a published stack of SOC 2 Type 2, ISO 27001, ISO 27701, GDPR, and HIPAA.

• Rapport publishes prices. Second Nature does not. Rapport lists a free 14 day trial, $19 per month Pro, $99 per month Business, and a quoted Enterprise tier. Second Nature is quote-only, and G2 reports a minimum average deal around $20,000 a year.

• Data location is the sharpest split. Second Nature stores customer data in a Google Cloud data center in the Netherlands. Rapport offers on-premise and private cloud hosting on Enterprise, so the data can stay inside your own environment.

• Scale favors Second Nature, control favors Rapport. Second Nature reports more than 1 million role plays and 50,000 trainees against Rapport's 610,000 role plays, while Rapport lets you bring your own characters, voices, brand, and even your own game engine build.

This is the closest comparison in the AI roleplay category, because both platforms show the learner a face. Rapport builds that face itself, in real time. Second Nature assembles a program around it. Everything below follows from that.

All pricing, customer, and security claims on this page were read from each vendor's own website. Vendors change these pages often, so re-verify before you quote anything in a board deck.

A platform's origin story is not marketing trivia. It predicts what the product is good at three years later, because the hardest engineering problem a company solves first tends to stay its moat.

Where Rapport came from

Speech Graphics was founded in Edinburgh in 2009 and spent a decade solving one problem: making a digital face move correctly from audio. That work shipped in blockbuster games, and it is why the avatar reacts instead of miming.

Rapport spun out in 2019, launched a self-serve platform in July 2024, and acquired Aquifer Motion in October 2025 to add full-body animation and cinematic camera tools. The company runs from Edinburgh with offices in San Francisco, Austin, Budapest, and Singapore.

Where Second Nature came from

Second Nature was founded around 2018 in Tel Aviv by Ariel Hitron and Alon Shalita, and took its first funding in 2019. The original product was Jenny, an AI sales coach you practiced a pitch with, and the company built outward from there into a full training program platform.

It raised a $22 million Series B in October 2025 led by Sienna VC, with Zoom taking part again after joining the Series A through the Zoom Apps Fund, bringing total funding to about $38 million. The team is roughly 45 people across Tel Aviv and New York.

Read those two paragraphs again, and the roadmaps write themselves. One company spent fifteen years on faces and is now building the program around them. The other spent seven years on the program and is now adding the faces.

Both platforms show the learner a face, so the usual avatar versus no avatar argument does not apply here. The real question is how each face is produced, because that decides what it can do while the learner is still talking.

What Second Nature does

• Runs four practice formats, not one: live interactive conversation, a webcam pitch recording that gets scored, a slide-based presentation, and a product demo with screen sharing.

• Offers video avatars and moving avatars, both marked as new on its own site, alongside the original Jenny persona the product started with.

• Scores on knowledge accuracy, conversational style, and soft skills, weighted 70 to 30 in favor of knowledge by default, and managers can change the weighting or add a manual evaluation.

• Returns feedback within about 45 to 90 seconds of the session ending.

• Scores what the learner does on screen against a recorded gold standard, so a rep can be marked on updating the CRM while handling the call.

What Rapport does

• Renders a 3D avatar live, with facial animation driven by the learner's audio at roughly 50ms, so the reaction lands mid-sentence rather than after it.

• Uses real-time rendering rather than generated or pre-recorded video, which is why the same avatar responds differently to the same words said two different ways.

• Lets you pick character, voice, language, and AI behavior, with characters from a built-in library, MetaHuman, Ready Player Me, or your own upload.

• Scores each attempt on competencies such as active listening, empathy, and clarity, and rolls attempts up into cohort analytics.

• Ships as a hosted link, an iframe, a web widget, a SCORM 1.2 package, or an Unreal Engine integration.

What this changes in practice

Generated video avatars are very good at looking like a person. Real-time animation is good at behaving like one. If your learners are rehearsing a message, the first is plenty. If they are learning to notice a reaction and change course, the timing of that reaction is the whole exercise.

Be fair about the trade. Second Nature's reviewers on G2 say the moving avatars are not perfect yet, and Second Nature acknowledges they are new.

Rapport has had fifteen years to get that one thing right, and correspondingly less time on the program layer around it.

There is a good chance your training business case opens with 70-20-10: seventy percent of learning from experience, twenty from other people, ten from formal training. It is the most quoted number in learning and development, and it is not a research finding.

The figures came out of work at the Center for Creative Leadership in the 1980s, where a group of executives were asked to recall how they had learned to do their jobs. It was a survey of memories, not a controlled study, and the numbers were never meant as a budget formula. Decades of repetition turned a rough recollection into a rule.

Worth checking before your next business case

The reason this matters for an AI roleplay purchase is that 70-20-10 is usually deployed to argue against formal training and for on-the-job experience. AI roleplay sits awkwardly in that frame, because it is neither.

It is repeated experience without the cost of learning on live customers, which is exactly the gap the model never accounted for.

Use numbers you actually own instead:

• Attempts per learner. Both platforms report it. Forty practice conversations before a first live call is a real change in behavior, and you can prove it happened.

• Improvement between first and last attempt. Rapport reports this by competency. Second Nature reports it against knowledge, style, and soft skills.

• Time to certification. If you certify before someone goes live, the days between hire and certification is a clean number your CFO already understands.

• The cost of the practice you replaced. Manager hours spent on live roleplay, at a real hourly rate, is the honest comparison.

A business case built on four numbers you can pull from your own dashboard survives procurement. One built on a 1980s recollection does not.

Every conversation runs on four layers at once. Sorting platforms by which layer they measure is faster than sorting them by feature list, and it is the only way to separate two products that both show you a face.

1. Context. Who is this person, what do they want, what is at stake, what happened last time.

2. Words. What you actually said, the questions you asked, the objections you handled, the framework you followed.

3. Tone. Pace, pauses, energy, whether you sounded certain or rattled.

4. Nonverbals. The face across from you, and whether you noticed it change.

Both cover context and words, and Second Nature covers them with more program machinery around them.

It's 70 to 30; the knowledge and style split is explicitly built to check whether the rep got the facts right, which is why it does well in certification programs where accuracy is the point.

Both handle tone, and here they diverge. Second Nature scores conversational style after the fact, in the 45 to 90 seconds following the session.

Rapport uses tone during the session, because the learner's audio is what drives the avatar's expression while they are speaking.

Layer four is the one to test in a demo rather than take on trust. Second Nature has video and moving avatars, so it is not absent from this layer.

The question is whether the face responds inside the conversation or performs alongside it. Ask both vendors to run the same scenario twice, once calm and once rattled, and watch what the face does.

This is the section where the two platforms are furthest apart, and it is the one most likely to decide your evaluation. Ask who builds scenario number 40, not scenario number one.

How Second Nature builds a scenario

Second Nature is fully self-serve, and this is its clearest strength. No technical expertise is required, and the tooling is built for an enablement team that wants to ship weekly.

• Upload a PDF, a PowerPoint deck, a Word document, or an audio recording, and build a roleplay from it.

• Or describe the scenario in plain text and the AI assistant builds the persona, the context, and the objection handling in minutes.

• Start from customizable prebuilt templates, or build from scratch in the course editor.

• Second Nature says teams can stand up full libraries of scenarios within days.

• Managers can adjust the AI scoring criteria and add their own manual evaluation on top.

How Rapport builds a scenario

Rapport is where you should ask the hardest questions on this topic, because the marketing site and the product documentation do not currently agree.

• The platform ships a scenario template library sorted by domain, covering learning and development, sales, and customer experience, and templates save to a personal library.

• The documentation states that scenario configuration fields are visible but not yet editable, with custom scenario creation listed as a roadmap item.

• Enterprise contracts include professional services and prompt engineering support, which is how bespoke scenarios get built today.

• Scenario fields, once editable, cover character, learner, and avatar roles, situation context, learner objectives, and the competencies to measure.

Be direct about this in your evaluation. If your enablement team needs to publish new scenarios every week without a services ticket, Second Nature is ahead today, and it is not close.

Get a written date from Rapport for self-serve scenario editing before you sign.

The counter-argument is real too. A smaller set of very high-fidelity scenarios that a specialist configures well beats fifty rough ones nobody finishes, and regulated buyers often prefer a services model precisely because it controls what gets published.

Both platforms produce reports. They answer different manager questions, and the difference is easiest to see side by side.

The honest summary: Second Nature measures whether a training program is working across a large organization.

Rapport measures whether a person handled a human moment. If you are certifying hundreds of people against a body of knowledge, the first is closer to the job. If you are changing how a smaller group behaves in the room, the second is.

This is where deals slow down. Both vendors say yes to LMS delivery, and only one of them publishes the details your IT team will ask for.

Two things to take from that table. Second Nature has the better answer for a large L&D team that already runs a corporate LMS and an identity provider, and it is the only one of the two that publishes score passback.

Rapport has the better answer when the data cannot leave your environment, or when you want the avatar inside your own application rather than inside a training portal.

The Netherlands detail is worth raising early. It is a real advantage for European buyers and a real question for United States federal, defense, and some healthcare buyers who need domestic residency.

Be straight with your security team about this one. Second Nature is ahead on certifications, Rapport is ahead on control, and those are two different kinds of safety. Knowing which one your reviewers care about makes the process shorter.

Second Nature security

• Publishes SOC 2 Type 2, ISO 27001, and ISO 27701, which is the privacy information management standard and is rarer than the other two.

• States GDPR, CCPA, and HIPAA compliance on its enterprise page.

• Stores customer data in a Google Cloud data center in the Netherlands, encrypted with AES-256 in transit and at rest.

• States clearly that customer data is never used for AI model training and remains isolated, and that it does not train large language models on customer data.

• Runs on Google Cloud and Microsoft Azure, with the language models hosted on Azure.

• Gaps to press on: there is no public trust center, no published data residency option outside the Netherlands, and no on-premise or single-tenant offering.

Rapport security

• States that your data stays yours, is encrypted, stored securely, and is never sold or used to train AI models.

• Publishes technical detail in the developer documentation, including HTTPS and TLS, certificate issuance, and encrypted WebRTC media over DTLS. See the Rapport cyber security details.

• Operates under UK GDPR and the Data Protection Act 2018 as Speech Graphics Ltd, registered in Scotland, with an outsourced data protection officer.

• Uses EU Standard Contractual Clauses with the UK Addendum for international transfers.

• Offers on-premise and private cloud hosting on the Enterprise tier, which is the strongest control available in this comparison.

• Gaps to press on: no SOC 2, ISO 27001, or HIPAA certification is published on the public site, and there is no trust center. Ask for the current security pack and any audit reports under NDA rather than assuming either way.

How to handle this in a real evaluation. If your procurement team treats SOC 2 Type 2 and ISO 27001 as a hard gate, Second Nature clears it today, and Rapport has a question to answer, so raise it in week one rather than week ten.

If your gate is where the data physically sits, the positions reverse: Rapport is the only one of the two publishing an option to keep it inside your own environment, and Second Nature's single Netherlands region is a fixed answer you either accept or do not.

One of these vendors publishes prices. The other does not. That is not a scandal, it is a go-to-market choice, and it changes how you run the evaluation.

Second Nature pricing

• No published figures at all. There is no pricing page and no tiers. Every path on the site leads to a demo booking.

• A free trial exists, and you can try sample roleplays without payment, but the length and terms are not published.

• G2 reports a minimum average deal starting around $20,000 for mid-sized companies and enterprises, and rates the cost perception at the top of its scale. That is the most credible public anchor available, and it is still third-party.

• Comparison blogs quote roughly $30 to $40 per seat per month. Both of the sites repeating that figure sell competing products, so treat it as a rumor rather than a rate.

• Reviewers say the model takes explaining. Several G2 reviewers describe needing multiple conversations before they understood how they were being charged.

Rapport pricing

• Trial: free for 14 days, 50 minutes of avatar conversation, no credit card.

• Pro: $19 per month, 100 usage minutes, 1 creator, up to 5 workspaces.

• Business: $99 per month, 500 usage minutes, 3 creators, unlimited workspaces, Rapport branding removed, role-based control, custom character uploads.

• Enterprise: quoted. Tailored usage packages, unlimited creators and workspaces, professional services, prompt engineering support, custom branding and reporting, dedicated CSM, and on-premise or private cloud hosting.

• Overage: $0.21 per minute above your allowance, against an effective included rate of $0.19 per minute.

How to compare a published rate against a quote

Rapport sells minutes. Second Nature sells an annual program. A 500-person certification program looks expensive on minutes and reasonable on an annual license. A team of 25 practicing occasionally flips that around completely.

Use this as your model:

• Take your learner count, the number of practice sessions per learner per month, and average session length in minutes.

• For a minutes-based quote, multiply all three, then multiply by the per-minute rate, and add expected overage at $0.21 per minute.

• For an annual license, ask what the number covers: named users or active users, whether managers and reviewers consume a seat, and what happens when headcount changes mid-term.

• Add the extras that sit outside the tier on both sides: implementation, professional services, and any scenario building you will not do yourself.

• Divide both totals by the number of completed practice sessions you expect in year one. That single cost-per-completed-session number is the only fair comparison.

Bring that model to both demos. Vendors respond very differently to a buyer who already knows their own volume.

Forget the category name for a moment. Find the row that matches what your team actually does badly.

If more than half your rows point one direction, you have your answer. If they split evenly, read the next section carefully, because you may be looking at two budgets rather than one decision.

Pick Rapport if

• The conversations you are training happen face-to-face, and noticing a reaction in the moment is the skill you are trying to build.

• Your security review starts with where the data lives, and an on-premise or private cloud option decides the deal.

• You want published pricing you can budget against without a sales cycle.

• The avatar needs to carry your brand, your character design, or a specific language and voice pairing.

• You are embedding the experience in your own product, site, or game engine build rather than in a training portal.

Pick Second Nature if

• You are running a large certification program and message accuracy is what you are measuring.

• Your enablement team wants to build and edit scenarios themselves, every week, from decks and documents they already have.

• Procurement treats SOC 2 Type 2 and ISO 27001 as a hard gate.

• You need the practice to live inside your existing LMS with scores flowing back automatically.

• You are training demos and presentations, not just conversations, and want screen sharing and slide walkthroughs scored.

• You are a BPO or an agency running separate programs for multiple clients under one roof.

When you genuinely need both

This pairing comes up more than you would expect, because the two products barely overlap in what they measure. A pharmaceutical company is the clearest example: the whole field force certifies on the launch message in Second Nature, and the smaller group who sit across from clinicians practices the harder conversation in Rapport.

A bank is another. The contact center certifies on policy and compliance at volume, and the relationship managers practice difficult client conversations with a face that reacts.

If that is you, split by audience rather than forcing one tool to do both jobs badly, run both on annual terms, and revisit at renewal once you have a year of completion and improvement data from each.

Score both platforms on the same ten criteria, weight them for your own program, and run the same nine questions on both demos. It takes an afternoon, and it removes most of the guesswork.

The 10-point evaluation scorecard

9 questions to ask on both demos

1. Build me a new scenario right now, from my own deck, using my product and my buyer. How long did that take, and who did it?

2. Play the same roleplay twice, once calm and once rattled, and show me what changed in the face, the response, and the score.

3. Show me the scoring rubric as an admin sees it. What can I change, and what is fixed?

4. What happens when the learner goes completely off script or tries to break the simulation?

5. Show me the exact LMS delivery path, including SCORM version, what data comes back, and which tier it requires.

6. Which region does my data sit in, is it used to train any model, how long do you retain it, and who are your subprocessors?

7. Price my actual volume: this many learners, this many sessions a month, this long each. Show the math and the renewal.

8. Show me the manager view for a team of 30 after two months, with real-looking improvement data, sliced by hire wave.

9. Run the same scenario in the second language my team uses, and let me judge the quality myself.

Ask questions six and nine in writing. Verbal answers on data handling and language quality do not survive a security review or a rollout.

Most teams that move are not unhappy with Second Nature. They move a specific audience, usually because the conversations that matter most stopped being about message accuracy and started being about how someone reacts.

A switch that goes well looks like this:

1. Export what you already have. Pull your scoring criteria, your persona definitions, and your completion data out before the contract ends. The criteria are the valuable part, because they encode what good looks like at your company.

2. Separate knowledge from behavior. Second Nature scores knowledge accuracy as the larger share by default. Decide which of your criteria are really knowledge checks, because those may be better served by a quiz than by a roleplay.

3. Rebuild your top five scenarios first. Not all forty. The five that account for most of the practice volume, so you can compare like for like against your existing baseline.

4. Run both for one quarter. Keep the certification program on Second Nature while the harder conversations run on Rapport. Compare completion rate, attempts per learner, and improvement between first and last attempt.

5. Confirm delivery early. Get the SCORM package into your LMS in week one, not week ten, and check that completion data lands where your reporting expects it.

6. Decide at renewal with data. By then you will know whether you have one program or two, and the answer is often two.

Going the other direction works the same way, and there is no shame in it. If your program is fundamentally about certifying a large group on a body of knowledge, the tool built for that is the right tool.

Want to see how your own scenario behaves with a face that reacts in real time? Try a live role play or book a demo and bring one real conversation your team keeps getting wrong.

Is Rapport a good Second Nature alternative?

It depends on which half of the job matters more to you. Rapport is a strong alternative when the skill is reading a person in real time, when the data has to stay in your environment, or when you want the avatar inside your own product.

It is a weaker alternative if you need a large self-serve certification program with LMS score passback, which is what Second Nature is built for.

What is the main difference between Rapport and Second Nature?

Both show the learner a face. Rapport renders a 3D avatar live and animates it from the learner's audio at roughly 50ms.

Second Nature offers video and moving avatars, recently added, inside a broader training program platform with course building, certification, and its own LMS.

Does Second Nature have a free trial?

Yes, and you can try sample roleplays from the site without paying first, but Second Nature does not publish the trial length or terms. Rapport publishes its trial: 14 days and 50 minutes of avatar conversation, no credit card.

How much does Second Nature cost?

Second Nature does not publish pricing. G2 reports a minimum average deal starting around $20,000 a year for mid-sized companies and enterprises, and comparison blogs quote roughly $30 to $40 per seat per month. Those blogs sell competing products, so get your own quote.

How much does Rapport cost compared to Second Nature?

Rapport publishes its pricing: a free 14-day trial with 50 minutes, $19 per month for Pro with 100 usage minutes, $99 per month for Business with 500 usage minutes, and a quoted Enterprise tier.

Because Rapport bills minutes and Second Nature quotes an annual program, compare them on cost per completed practice session.

Which one has better avatars?

Test it rather than take either answer on trust. Rapport's animation is driven live from audio and is the thing the company has spent fifteen years on.

Second Nature's video and moving avatars are newer, and its own G2 reviewers describe them as not perfect yet. Run the same scenario on both and judge the reaction, not the still image.

Is Second Nature SOC 2 compliant?

Second Nature publishes SOC 2 Type 2, ISO 27001, and ISO 27701, along with GDPR, CCPA, and HIPAA, and states that customer data is never used to train models.

Rapport does not publish those certifications and operates under UK GDPR and the Data Protection Act 2018. Ask both vendors for their current security pack in writing.

Where is my data stored?

Second Nature stores customer data in a Google Cloud data center in the Netherlands, and no alternative region is published. Rapport offers cloud, on-premise, or private cloud hosting on the Enterprise tier.

If domestic data residency is a requirement, this is the question to ask first.

Can you use Rapport and Second Nature together?

Yes, and some organizations should. The common split is Second Nature for the large certification program and Rapport for the smaller group whose hardest conversations happen face-to-face.

Split by audience rather than trying to make one platform cover both jobs.