Full session localisation for non-English roleplays Localisation
When you run a simulation in a language other than English, the live session, microphone check, and post-session feedback now follow that language end to end, including score explanations, so learners stay in one language from start to finish.
Studio admins can configure end-of-session surveys for their learners, with rating, NPS, and free-text questions. Responses appear in session history, analytics, and CSV exports so programme owners can measure usefulness and gather ideas for new scenarios.
After a simulation, learners can tap Agree or Disagree on individual feedback cards so teams can see which coaching points landed. Studio admins see reaction summaries alongside survey results in reporting and exports.
The site now includes a Customer Stories section with video interviews and case studies from organisations using Real Talk Studio, so prospects and teams can hear how others practise difficult conversations before they sign up.
Organisation super admins now receive an email when post-session moderation flags learner messages, including the triggering content and a link to review the session, so compliance teams can act quickly without polling the admin hub.
Real Talk Studio now supports Korean across the platform interface, roleplay creation, and practice flows, with matching voice options so Korean scenarios can run end to end.
Pause and resume live sessions Conversation Engine
During avatar and audio-only simulations you can pause the session and pick up where you left off. The timer stops while paused, and live connections are torn down so you are not charged for idle time.
Before you begin a practice simulation you work through a short warm-up: preview how the conversation might open, consider a pivotal moment, choose a skill to focus on, and rate your confidence. The flow runs when you start practice (including demo sessions) so you enter the roleplay prepared rather than jumping straight to the microphone check.
New Security, Trust Centre, and Responsible AI pages give procurement and compliance teams a clear view of our ISO certifications, data practices, and how AI is governed in practice sessions. An on-brand certification banner highlights our ISO 27001, 9001, and 42001 credentials on the homepage.
New Security, Trust Centre, and Responsible AI pages give enterprise buyers one place to review ISO 27001, ISO 9001, and ISO 42001 certifications, policies, and how AI is governed in practice sessions. The homepage highlights ISO certification for teams evaluating the platform.
Real Talk Studio now supports Polish end to end: the interface is available in Polski, and you can create and run Polish roleplays with speech recognition tuned for the language.
Studios can enable preparedness tracking so learners rate how ready they would feel before a first attempt and again after the session, with repeat runs able to reference their previous score. Studio admins see before and after scores in session history, analytics, and CSV exports.
Studio admins can create cohorts, assign members, and filter session history and analytics by cohort so reporting matches how programmes are organised.
Live roleplay sessions now keep practice focused on the scenario. If your messages drift too far off topic you see a warning in the transcript, and the session ends if off-topic turns continue, so time in the simulation stays aligned with the learning objective.
After harassment practice simulations, learners complete a guided Take Action path before feedback: recognise what happened, choose who they would tell (line manager, HR, whistleblowing, or keep it private), and record their reflection, supporting compliance rehearsal without jumping straight to scores.
During live simulations you get a clearer progress view (elapsed time and target duration), an orientation strip to open scenario context when you need it, and a collapsible side panel on larger screens. Mood is easier to read at a glance with a new ring indicator, achievements show locked and unlocked states more clearly, and ending a session now asks for confirmation so you are less likely to leave by mistake.
Guided steps for AI roleplay creation Roleplay Builder
The AI-assisted roleplay builder now follows a clearer step-by-step flow, with dedicated screens for templates and authoring instead of juggling several panes at once, so you can move from idea to a ready-to-review scenario with less friction.
Organisation admins reviewing interview and studio sessions can see persona details and non-verbal presence alongside session history and CSV exports when that data was captured, supporting fuller coaching and compliance review.
The application now supports LemonSlice and LiveAvatar only. HeyGen Streaming, Simli, and their npm packages have been removed. IT firewall guidance and environment variables were updated accordingly.
Fast track is now the default way into a simulation, aligned with current branding and styling. Session start includes the same microphone and camera checks as other flows where applicable, so setup stays predictable before you begin.
The platform continues to broaden language coverage — including French in the interface set alongside existing locales, additional locales where relevant, and expanded voice language options so more scenarios match your learners’ languages.
Sessions now use a single dark theme end to end — a deliberate branding and UI refresh so live simulations, completion, and shared feedback share the same look and feel instead of mixing light and dark surfaces.
Post-session feedback has been reworked: the completion flow is built around the transcript, with a clearer tabbed layout for reviewing outcomes. Shared feedback matches the new design system so learners and stakeholders see the same polished experience after a simulation.
MCP support lets teams and tools create roleplays from connected clients, with documentation and on-site copy explaining how Real Talk Studio integrates with MCP-compatible environments.
Simulations can now run in non-verbal mode — practise reading cues and responding without relying on spoken dialogue, with clear messaging during camera and session setup where streaming behaviour differs from standard voice sessions.
Every simulation session now receives an automatic realism score — a measure of how closely the AI character’s behaviour mirrors a real conversation. Admins can review scores across sessions to spot which scenarios are performing well and where prompts may need tuning.
The homepage and practice pages now feature preview videos and direct links to example simulations, so visitors can see the platform in action before signing up.
The platform uses LiveAvatar for realistic streaming avatars (via LiveKit). Legacy HeyGen Streaming and Simli integrations have been removed from the product codebase.
Custom avatar image generation now produces more neutral, natural facial expressions and makes better use of screen space — characters look more lifelike and less artificial.
Conversation flow is now generated when a roleplay is created rather than during the session, meaning faster starts and more consistent scenario delivery.
AI characters now express emotions with more subtlety and variation. The mood system has been refined to avoid rigid or predictable responses, producing more authentic conversational dynamics.
A major new simulation mode designed specifically for audit preparation — practice responding to auditor questions in a realistic, pressure-tested environment.
A new simulation type for practising presentations and pitches. Deliver your content to an AI audience that responds with realistic questions and feedback. Includes moderation controls and post-session user feedback.
New tooling for generating promotional videos directly within the platform, replacing the previous static video assets with dynamic, up-to-date content.
The pricing page has been redesigned with clearer tiers including a new free plan, making it easier for teams to evaluate Real Talk Studio before committing.
A new scoring dimension that measures how well the learner followed relevant policies during the simulation — available in session completion, roleplay setup, and review screens.
A new pre-built persona type optimised for customer-facing scenarios, with an existing customer profile that adapts behaviour for service and sales simulations.
AI character personalities are no longer static — they now shift and respond dynamically during sessions, creating more authentic and unpredictable conversations.
Comprehensive product documentation is now available at docs.realtalkstudio.com, covering setup guides, network requirements, and platform capabilities.
Post-session reviews now include detailed feedback on the learner’s speaking performance — pacing, clarity, and delivery alongside the existing content-based scoring.
A new “assisted mode” replaces manual turn-taking, automatically managing conversation flow so learners can focus on what they’re saying rather than when to speak.
A brand new way to run simulations — choose between avatar-based video sessions or audio-only mode at session start. Audio-only is ideal for phone-call simulations, lower-bandwidth environments, or when you want to focus purely on the conversation.
Cartesia has been added as a text-to-speech provider, delivering more natural-sounding voices with expanded language support and dynamic emotional expression.
Improved connection stability with WiFi status indicators, automatic session recovery on connection drops, and graceful handling of browser closures — your session progress is now saved even if you close the tab.
