Presence

Inspiration

People spend a great deal of time preparing what they will say before an interview, meeting, or presentation, but very little time preparing how they will actually appear on camera.

Poor lighting, awkward framing, distracting backgrounds, or an uncomfortable camera angle can shape a first impression before the conversation even begins.

Most camera-check tools stop after confirming that a webcam works, while most Skin AI applications focus on beauty, skincare, or cosmetic recommendations.

We wanted to explore a different question:

Could Skin AI help people prepare for important on-camera moments without judging the person in front of the camera?

That question became Presence.

Our guiding principle is simple:

Do not evaluate the person. Improve the environment they can control.

---

Building Presence

Presence guides users through a short preparation loop before interviews, online meetings, presentations, and profile photos.

Real-time analysis runs locally using MediaPipe to evaluate face position, framing, head orientation, brightness, and background complexity. Once the user explicitly requests a report, a single optimized still image is analyzed using YouCam Skin Analysis, Skin Tone Analysis, and Photo Lighting.

One of the biggest product decisions came from working with the YouCam APIs.

At first, we treated the returned signals as candidates for a single overall score. As we explored the APIs, we realized that each feature represents something fundamentally different. Skin Analysis measures individual skin concerns, Skin Tone Analysis provides color-related information, and Photo Lighting generates an AI-enhanced preview rather than a numerical measurement.

That realization changed the product.

Instead of creating another appearance score, we redesigned Presence around Camera Readiness.

The headline score measures only controllable camera conditions such as composition, lighting, and visual focus. Skin concern measurements like texture, pore, redness, and dark-circle scores are intentionally excluded.

MediaPipe tells Presence where the face is.

YouCam helps Presence understand how the final captured image renders.

Presence combines those two layers to decide what the user should change next.

Rather than exposing raw AI outputs, Presence translates the analysis into practical actions, such as adjusting camera height, changing the direction of a light source, reducing harsh shadows, or simplifying the visible background. After making changes, users capture another real image to verify the improvement instead of relying on an AI prediction.

---

Challenges

One of our biggest technical challenges was combining three independent YouCam workflows into a single, reliable user experience.

Each feature requires image upload, task creation, polling, and response normalization, while every API returns a different response structure. Some analyses can succeed while others fail independently, and Photo Lighting behaves differently because it generates an image rather than a numerical result.

To solve this, we built a normalization layer that treats every feature independently, executes requests concurrently with Promise.allSettled, and explicitly reports whether each result is available, unavailable, failed, or running in demo mode.

We also wanted to preserve both privacy and responsiveness.

Real-time camera guidance stays entirely on the user's device. Only one optimized still image is uploaded after the user explicitly chooses to generate a report. Previous captures remain in the browser for before-and-after comparison instead of being stored on a server.

Another challenge was ensuring that Presence never implied judgments about the person.

The product does not estimate attractiveness, professionalism, personality, confidence, or hiring potential. Every recommendation focuses only on things the user can actually change before an important online moment.

---

What we learned

The biggest lesson was that the value of an AI API is not simply the information it returns.

A product must also decide what that information should mean, what it should never imply, and what action it should support.

Working with the YouCam APIs taught us that different analysis results should not automatically be merged into one score. Preserving the meaning of each signal produced a product that is both more honest and easier to understand.

We also learned that explicit partial results are more trustworthy than silent defaults. Instead of inventing realistic-looking fallback values, Presence clearly communicates when information is unavailable while continuing to provide useful guidance.

Most importantly, we learned that Skin AI can create meaningful value outside traditional beauty and skincare when paired with a clear ethical boundary.

Presence uses AI to help people improve their camera environment—not to judge the person in front of the camera.

---

What's next

Our next goal is to make each scenario—interviews, meetings, presentations, and profile photos—change not only the wording of the guidance but also the recommendation priorities themselves.

We also plan to validate our readiness model through user studies across different devices and lighting environments, explore optional apparel color guidance and Fashion VTO previews, improve presentation feedback without making personality inferences, and expand Presence into preparation tools for universities, coaching services, and remote-work organizations.

Our long-term vision is simple:

Make Presence the preparation layer people use before every important on-camera moment.

Built With

• boundra
• mediadevices
• playwright
• svelte
• sveltekit
• tailwind
• typescript
• vercel
• vitest
• webrtc