DEV Community

Shivam Chaturvedi
Shivam Chaturvedi

Posted on AI-assisted

mocktwo - Practising High-Pressure Conversations with Presage SmartSpectra

Presage SmartSpectra SDK Challenge Submission

This is a submission for the Presage SmartSpectra SDK Challenge

What I Built

Have you ever walked into an important client meeting, interview, presentation, or negotiation feeling prepared, only to become nervous when someone challenges you?

Your heart starts racing, your breathing changes, and suddenly it becomes harder to think clearly and communicate confidently.

Most people prepare for these situations by reading notes or practising in front of a mirror. But that doesn't prepare you for the pressure of a difficult conversation.

That's why I built mocktwo, an AI-powered conversation rehearsal platform that helps people practise high-pressure situations and understand how their bodies respond under pressure.

Whether you're preparing for a client escalation, an important sales pitch, a job interview, a difficult workplace conversation, or an executive presentation, you can:

  1. Choose a real-world scenario.
  2. Select an AI persona with a specific personality and communication style.
  3. Start a face-to-face conversation with an AI counterpart powered by Tavus AI.
  4. See your physiological signals measured in real time using Presage SmartSpectra SDK.
  5. Review your performance, understand your pressure moments, and practise improving your composure.

The idea is simple: don't just prepare what you're going to say. Practise handling the pressure that comes with saying it.

mocktwo is built using Electron.js, with support for desktop and web experiences.

Demo

Want to try it yourself? Download the desktop application:

Download mocktwo for macOS

Code

The complete source code is publicly available:

GitHub Repository: mocktwo

Run locally

git clone https://github.com/shivam-chaturvedi/mocktwo.git
cd mocktwo
npm install

cp .env.example .env
Enter fullscreen mode Exit fullscreen mode

Add your credentials to .env:

SMARTSPECTRA_API_KEY=your_presage_api_key
TAVUS_API_KEY=your_tavus_api_key
Enter fullscreen mode Exit fullscreen mode

Then run:

npm run dev
Enter fullscreen mode Exit fullscreen mode

How It Responds to Human Signals

This is where mocktwo becomes different from a regular AI roleplay application.

It doesn't just listen to what you say. It also helps you understand how your body responds while you're having a difficult conversation.

1. Practise with an AI counterpart using Tavus

First, choose a scenario. For example, imagine you have an unhappy enterprise client who is threatening to cancel a contract.

You select the scenario, choose a persona, and start the rehearsal.

We're using Tavus Conversational Video AI to power the AI counterpart. Instead of interacting with a text-based chatbot, you have a face-to-face video conversation with an AI persona that responds to your answers and challenges you based on the selected scenario.

It can question your decisions, push back on your arguments, and make the conversation feel more realistic.

2. Measure physiological signals using Presage SmartSpectra

While the conversation is happening, Presage SmartSpectra SDK uses your camera to capture optical signals and measure physiological indicators such as:

  • Pulse rate: How your heart rate changes during the conversation.
  • Heart Rate Variability (HRV): Variations in the time between heartbeats.
  • Breathing rate: How your breathing pattern changes.
  • Blinking and talking activity: Additional signals that help provide context to the session.

Before a rehearsal, mocktwo also provides a calm baseline calibration so that measurements can be compared against your own resting readings.

No smartwatch or external wearable is required.

3. Understand your pressure moments

After the rehearsal, mocktwo generates a detailed session review.

It combines physiological measurements with the conversation transcript and recording to help you understand:

  • When your pulse increased during the conversation.
  • How far it moved from your resting baseline.
  • How long it took to return closer to baseline.
  • Which parts of the conversation coincided with noticeable changes.
  • How much time you spent speaking versus listening.
  • Your speaking pace and other available session metrics.

The application calculates a Composure Score using pulse control and recovery measurements. Pulse control contributes 60%, while recovery contributes 40%.

It also categorizes the session into understandable levels such as Calm, Mild, Moderate, and High, based on the implemented scoring thresholds.

4. Review, learn, and improve

The Analytics and Coaching sections bring everything together.

The Pressure Timeline displays your pulse throughout the conversation, with markers highlighting significant increases. Selecting a marker takes you to the corresponding moment in the recording and transcript.

You can review the AI coach's feedback, identify moments where you rushed your response or struggled with an objection, and practise personalized drills.

Your previous sessions and progress are also available, making it easier to compare how you respond to similar situations over time.

The goal isn't to eliminate nervousness. It's to become more aware of your responses, practise handling challenging conversations, and improve how you communicate under pressure.

Prize Categories

I'm submitting mocktwo for both categories.

Best Thoughtful Real-World Use

High-pressure conversations are part of everyday professional life. People face difficult clients, unexpected questions during interviews, demanding negotiations, and presentations where the stakes are high.

mocktwo gives them a way to practise these situations before they happen.

By combining conversational AI with physiological feedback, it helps users go beyond rehearsing answers and understand how they respond when challenged.

Most Creative or Delightful Use

The interesting part of mocktwo is bringing two technologies together:

  • Tavus AI creates interactive video counterparts for realistic conversations.
  • Presage SmartSpectra SDK provides camera-based physiological measurements during those conversations.

Connecting the conversation with the user's physiological signals creates a different kind of rehearsal experience. Users can revisit a difficult moment, see what was happening in their measurements, and understand the conversation from another perspective.

Consent and Data Handling

mocktwo is designed around explicit user consent and personal data privacy.

  • Camera and microphone access are requested when the user starts the relevant experience.
  • Presage SmartSpectra processes camera frames locally for physiological measurement. Raw frames used for this processing are not stored or sent to Presage.
  • SmartSpectra SDK telemetry is disabled.
  • During live AI conversations, camera and microphone streams are sent through the encrypted WebRTC connection used by the Tavus experience so the AI counterpart can interact with the user.
  • Physiological measurements, session recordings, and transcripts are stored locally on the user's device.
  • Users can delete their saved recordings and associated session data.
  • Physiological monitoring is active only during user-initiated calibration and rehearsal.

mocktwo is an educational rehearsal and coaching tool. It is not intended to diagnose, treat, or prevent medical conditions.


mocktwo is built around one idea: preparation shouldn't stop at knowing what to say. You should also get the opportunity to practise how you respond when the pressure is real.

Top comments (0)