Léo Desrumaux was only a 16-year-old kid when his life changed. “I moved to the US as a foreign exchange student. I didn’t speak a word of English. I really didn’t have a lot of confidence in myself. I was a little bit lost,” he tells me in a video interview. “Then I found myself through boxing.” Sixteen years later, Desrumaux discovered digital fitness and embarked on a quest to recreate the boxing training experience at home. The result is Growl, a device that turns a punching bag into an AI-powered trainer capable of seeing you, stopping your punches, and directing you through a workout using multiple sensors and three-dimensional cameras.

The design process, Desrumaux explains, started with the coach experience. “Our premise is recreating that physical connection with the coach; then how do you bend the technology to your will so that it really augments the experience that you want?” he says. Desrumaux believes that devices like Peloton started from the hardware—the traditional static bike—then added content on top of it. In their case, they thought first about how the coach should work and then built the bag from there.

Growl is centered on an augmented reality coaching experience. “A lot, if not everything, that we use is through light,” Desrumaux says. The image of the coach comes from a 4K high-brightness projection model that recreates that image. The projector is positioned on the top of the bag’s frame at an angle, so you’re not going to break light. The frame also houses the multidirectional time-of-flight infrared sensors that, like LiDAR, project laser light patterns on your body.

“When you come in and touch the bag, you break that light pattern, and it knows where you are interacting and when you were touching the bag,” Desrumaux explains. The cameras measure impact force, speed, and accuracy, capturing data each time the user makes contact.

Additionally, the time-of-flight sensors are used to monitor positioning and provide a responsive training environment from anywhere in the room. The cameras are located on the wings of the boxing frame, top and bottom, so regardless of your height, left or right-handedness, it captures and recreates your skeleton in real-time, Desrumaux says. If you are lifting weights and doing other exercises, Growl knows exactly how you are doing it.

The bag itself is a different form of a traditional heavy boxing bag. Instead of hanging freely, Growl’s reinforced boxing bag has a frame backplane that attaches to a wall in your home. The surface is made of leather and inside is foam. “You can shoot it with bullets, and it’s not going to break because it’s like punching a mattress against your wall,” he says. This will guarantee its durability, he claims, which is a main concern for consumers because, obviously, hitting your expensive piece of hardware won’t quite work if it can’t sustain the abuse long term.

A new type of fitness user experience

The other fundamental part of Growl is the software. The AI component runs in the GPU located in the arm above the bag. Desrumaux describes it as “the brain of the system.” He says it uses custom machine learning algorithms that have been trained to recognize proper form, detect flaws, and deliver corrective feedback instantaneously. “Growl isn’t just counting your reps or measuring power,” Desrumaux tells me. “It’s actively analyzing your entire workout and adjusting the training program accordingly.”

The user interface is projected onto the bag: A life-sized coach offers real-time advice, data visualization, and progress tracking, as well as telling you how to move. The coach is extremely interactive. When you are training for boxing, for example, he is wearing focus mitts, offering targets to hit and providing instant feedback.

The system compares your movements to those of this virtual coach, identifying incorrect techniques like dropping your guard. “During or at the end of the class, we’ll provide cues on what you could improve and even include videos on how to do it correctly,” he says. Desrumaux says the cameras use depth-sensing technology similar to that found in autonomous vehicles or your iPhone’s Facetime camera, giving Growl the ability to assess positioning and technique with precision.

Right now, the Growl app offers pre-programmed content with cues from the AI to adjust your form, motivate you, or keep you in rhythm. In the future, the company wants to provide with a library of personalized routines that adjust based on user performance. Desrumaux mentions that “the system creates dynamic workouts; it reads your fatigue, how well you’re doing, and adapts.” Unlike traditional fitness apps that rely on pre-programmed content, Growl uses machine learning to generate a unique training experience every session.

A typical Growl training starts with a quick warm-up that primes the user for the workout. The AI observes the warm-up in real-time, adjusting the exercise if necessary based on user feedback or detected stiffness. Once the core session begins, the bag becomes an active partner, issuing commands such as punch combinations or footwork drills.

All the system’s sensors combine to provide instant feedback, with the AI making on-the-fly adjustments to ensure proper execution. “If your stance is off, Growl will correct you,” Desrumaux says. “If your strikes are losing power, it’ll prompt you to adjust.” Sessions conclude with a cooldown phase that is also fully adaptive, aimed at optimizing recovery.

A new kind of digital workout

Growl is positioned differently from existing digital fitness products like Peloton or Tonal. While Peloton offers an aerobic-heavy, bike-based experience, and Tonal leans on resistance-based strength training, Growl seeks to fill a niche focused on combat sports and high-intensity interval training (HIIT).

Desrumaux tells me that their goal was to create a system that provides the intensity of a physical boxing gym but makes it accessible to individuals at home. “Boxing is something you can’t fake—it’s about precision, speed, power. Growl captures all of that,” he says. The UX aims to create a sense of engagement similar to Peloton, but in a way that offers direct, physical interaction with a piece of equipment that responds to you in real-time. “Peloton is a very static experience—you just sit and cycle—but with Growl, you’re up on your feet, moving, hitting something that moves with you. It’s about immersion and that real feeling of interaction,” he says.

Pre-orders are not yet open, but you can join the waiting list now. The company says they anticipate hardware price to be at $4,500, with a monthly subscription of $60 that covers an unlimited number of family users. The subscription includes access to the personalized workout library, detailed analytics, and live training sessions with professional boxers.

Desrumaux tells me that Growl is intended to be a long-term fitness companion, and that the system will continue to get smarter with each update. He says his company has plans to integrate third-party wearables to gather more detailed biometric data and improve the AI’s ability to adapt based on physiological indicators like heart rate variability and sleep quality.

The company also plans to introduce new training modes aimed at different skill levels, from beginners to competitive athletes, enhancing the scope of Growl beyond the casual fitness user. In the future, the Growl team is also planning to make it even more interactive by rendering a sparring partner in 3D. It all sounds like a great idea with lots of potential for good health and eliminating stress. I used to love boxing training so I would be really tempted—if I could only muster the energy to open a browser tab and hit buy rather than having another siesta.