Tensity

UX DESIGN

STRATEGY

Overview

In 2024, the integration of generative AI into products and services marked a significant shift in how consumers interact with technology. For more detailed research on identifying needs within the integrated fitness landscape, take a look at my report here.

This case study outlines the development of Tensity, a speculative AI-powered fitness solution designed to deliver personalized training recommendations. The Tensity system includes smart resistance bands that measure workout metrics and utilizes your smart device’s camera for motion tracking during workouts. This enables the Tensity app to provide real-time feedback during each workout.

Role: Solo project (research, UX/UI, visual design)

Duration: 10 weeks

Goal: To develop an affordable and low-footprint AI-driven home fitness system

Note: imagery for this project was created using generative AI with additional processing. Read on for more on the image generation process.

Understanding the market

To gain a deeper understanding of the current market, I performed a competitive audit on 5 leaders in the connected fitness space. These included high-end, all-in-one systems with low to no portability (Tonal, Tempo), a TV-connected camera designed for use with free weights (Peloton Guide), and portable all-in-one solutions with fewer features (LIT Axis, TRX).

Beginning with Tonal, Peloton, and Tempo, it quickly became clear that there was a need for more portable and affordable fitness solutions (the cost for these three ranged from $959 to $6,153 over a three-year period).

Zeroing in on portable fitness solutions, I analyzed LIT Axis and TRX, two leaders in the resistance band and suspension strap market. Although these options were more affordable, they lacked advanced functions such as AI form tracking and real-time feedback.

I wanted a solution that would provide the premium experience of a wall-mounted system like Tonal, but realized that the inclusion of a smart screen would compromise affordability.

Solution

I proposed a fitness solution that integrates a user’s existing smart screen into the workout system. The system includes wall brackets that hold a phone or tablet, alongside resistance bands and suspension straps that clip into adjacent brackets. For greater portability, the bands and straps can also attach to existing anchor points.

Tensity leverages the smart device’s front-facing camera for AI-powered motion tracking, while embedded sensors in the resistance bands measure the amount of force applied during a workout. This data is used to provide real-time feedback during each workout, and includes metrics such as rep count, form accuracy, and heart rate.

Tensity also features a generative AI personal trainer that answers common questions, provides progress updates, and generates custom training plans tailored to the user’s workout data. This creates a personalized and interactive fitness experience without the need for expensive equipment.

Site map & wireframes

To begin the design phase, I developed a site map to establish the app’s structure, focusing on intuitive navigation and content hierarchy. Using this information architecture, I created low-fidelity wireframes in Balsamiq, which allowed me to quickly create new design iterations.

Logo and style guide

The logomark draws inspiration from the resistance bands, incorporating the triangular handles and thin strap into its design. The color palette combines the energy of the orange with the stability of the blue to represent the balance between movement and control in fitness.

Image generation

To create the look and feel of a fitness app, I recognized that I would need to generate realistic photography. While proficient with Midjourney, I was excited to experiment with new image generation platforms.

I tried two additional platforms: Stability.ai’s browser-based platform DreamStudio, which uses Stable Diffusion XL v1.0 (SDXL), and Stable Diffusion 3.5L via the SwarmUI, a GPU-powered platform. Here are my biggest takeaways:

Detailed prompts are essential. Like most image generators, Stable Diffusion produces the most accurate results when the image prompt is highly detailed.

Results vary by platform. Each image generator produces entirely different results. I found that Stable Diffusion 3.5L had the best adherence to the prompt, but often produced nonsensical images. DreamStudio’s results looked realistic but struggled with human anatomy. Midjourney’s results looked less realistic overall, but reliably generated images of people with accurate anatomy, such as the correct number of limbs and fingers.

Ultimately, I used a combination of the three, often taking an image generated with one and using it as the reference image for another.

Some images turned out better than others.