Fortell AI Hearing Aids Review: A New Approach to Hearing in Noise

Increasingly, premium hearing aids from the world’s leading manufacturers promise better hearing in background noise through AI-driven processing and deep neural networks (DNNs). Fortell is making a more specific—and potentially more consequential—claim: its hearing aids can separate speech from competing voices and environmental sounds in real time, then process those sources differently based on what they are and where they originate.

If true, that could be a huge boon for people with hearing loss who are struggling to communicate in places like a restaurant, where background noise is not always the hum of an air conditioner or the clatter of dishes. Often, the “noise” is other people talking in the same frequency range as the person you are trying to follow. Conventional noise reduction can make the overall scene more comfortable, and directional microphones can emphasize sound from a chosen direction. Fortell says its Spatial AI goes further by separating incoming audio into different sources and treating wanted speech differently from competing speech and nonspeech noise.

Quick Take

Fortell says its Spatial AI processing tackles the problem at a more fundamental level. Instead of merely classifying the environment and adjusting a traditional processing system, the company says its neural network analyzes sound from four microphones across the pair, separates incoming audio into different sources, and then amplifies or attenuates those sources according to where they are and whether they appear to be important speech.

Fortell AI 2.0, the company’s first major software upgrade, launched on August 26, 2026 as a free update for new and existing users. Fortell AI 2.0 retrains the DNN on a larger collection of real-world recordings and improves several signal-processing algorithms without requiring new hardware. In Fortell’s testing across 80 speech-in-noise scenes, AI 2.0 increased SNR benefit by an average of 2.13 dB over the original model, with a best-case improvement of 4.1 dB in one competing-talker condition.

Fortell AI Hearing Aids are rechargeable prescription RICs for adults with mild to severe hearing loss. They provide four listening modes—All Voices, Front Voices, Conventional, and Live Music—and reportedly deliver at least 24 hours of battery life with AI processing continuously active.

The hearing aids cost $6,800 per pair, including 5 years of care, a 3-year warranty, loss-and-damage coverage, and a 100-day trial. Direct streaming and app support are currently limited to compatible Apple devices; Android and telecoil support are not available, although Fortell says it will add Auracast through a software update.

Fortell’s AI 2.0 findings are encouraging, but HearingTracker and the independent HearAdvisor lab are still conducting a hands-on review and standardized testing of the hearing aids. We will add the results, SoundScore, pros and cons, and final assessment as soon as we get them.

Why Fortell developed new hearing aids and where you can get them

Co-founder and CEO Matthew de Jonge has described Fortell as an attempt to break from decades of incremental hearing aid development. The company’s origin story is personal: he watched his grandparents become increasingly disconnected from family conversations despite owning expensive hearing aids. They could hear that people were talking but could not reliably follow the conversation when several voices competed.

That experience pushed Fortell toward one of hearing science’s best-known challenges: the “cocktail party problem,” or the ability to follow one talker in a noisy room full of other talkers. Fortell spent more than 5 years developing its system, first demonstrating the processing through large external computers before shrinking it into a wearable hearing aid.

How Fortell’s Spatial AI works

Many of the world’s best hearing aids now use artificial intelligence, but “AI hearing aid” can describe very different technologies. One common approach is environmental classification: a machine-learning model identifies various listening situations like a restaurant, car, quiet room, or music setting, then adjusts directionality, noise reduction, compression, and other parameters accordingly.

Fortell says its Spatial AI approach operates deeper in the signal-processing chain. Four microphones across the pair of hearing aids capture timing, level, spectral, and spatial information. A deep neural network then attempts to separate important speech from interfering sounds before presenting the processed scene to the wearer.

The company built a dedicated machine-learning coprocessor chip for this task. Fortell says it can perform up to 100 billion operations per second, offers roughly 235 times the computing capacity of a conventional hearing aid processor, and maintains end-to-end latency (the delay between sound entering the microphone and exiting the receiver/speaker) below 10 milliseconds.

The practical questions are whether source separation consistently delivers better performance in varied real-world scenes, keeps voices and environmental sounds natural, avoids objectionable delay or artifacts, and does so without excessive battery drain.

What changed with Fortell AI 2.0

The original Fortell model—now called AI 1.0—relied heavily on synthetic training data: computer-generated acoustic scenes that can be produced at enormous scale. AI 2.0 adds substantially more audio captured in places where hearing aid wearers often struggle, including restaurants, cafés, construction sites, trains, buses, cars, and reverberant offices and living spaces.

Real listening environments contain overlapping and moving talkers, reflective surfaces, clattering objects, and sound levels that change from moment to moment. Fortell’s premise is that training its model on more of this acoustic messiness will help it perform better when wearers encounter similar conditions.

The company also revised several conventional signal-processing algorithms. Fortell says the combined changes are intended to improve two related qualities: speech clarity in noise and the naturalness of voices. That second goal is important because increasingly aggressive speech separation—which usually emphasizes higher frequencies—has limited value if the desired voice becomes robotic, clipped, or distracting.

Controlling the Fortell hearing aid and its four listening modes

The Fortell app has four listening modes you can choose from:

• All Voices is intended as the everyday AI mode. It reduces background noise while preserving voices around the listener, making it better suited to group conversations where several people matter.
• Front Voices is the most aggressive speech-in-noise setting. It focuses on speech in front of the wearer while suppressing competing talkers from the sides and rear. This mode underlies Fortell’s strongest clinical claims and is best suited to noisy situations in which the important person can be positioned in front.
• Conventional turns off Fortell’s distinctive AI source-separation processing and operates more like a traditional premium hearing aid, using wide dynamic range compression (WDRC), conventional noise reduction, and adaptive directionality.
• Live Music is designed for concerts and performances, where strong speech-focused processing could interfere with music’s natural dynamics and character with distortion and acoustic artifacts.

Fortell deliberately requires the wearer to select Front Voices rather than switching into it automatically. That's because the processing is designed to suppress side and rear talkers, meaning it could remove someone the wearer would like to hear. Front Voices can't read your mind; it may be less useful when several important speakers surround the listener or when an unwanted talker is directly behind the desired speaker (e.g., when driving with someone in the backseat).

If you don't like reaching for the app, the left hearing aid controls volume, while the right changes programs. The iOS app adds mode selection, independent left/right volume, a 3-band equalizer, streaming controls, battery information, messaging, remote adjustments, and firmware updates. The Fortell app is not currently compatible with Android phones.

Battery life, connectivity, and remote care

Fortell reports more than 24 hours per charge with continuous AI processing, approximately 6 hours of use from a 30-minute quick charge, and more than 72 hours of additional use from the portable USB-C charging case. It does not support Qi wireless charging.

As previously noted, compatible iPhones, iPads, and Macs can stream directly to the hearing aids, and the app supports real-time remote fitting, adjustments that an audiologist can send for later installation, and over-the-air software updates. Android is not currently supported.

The lack of a telecoil may matter to people who regularly use hearing loops in theaters, lecture halls, places of worship, or public-service counters. Planned Auracast support could eventually provide access to compatible public broadcasts—although Auracast transmitters are growing in popularity but still rare.

Fortell’s clinical evidence for hearing aid performance

For its AI 2.0 white paper, Fortell compared the new and original software using the same hearing aid hardware and the same 80 KEMAR manikin-recorded speech-in-noise scenes. Half contained environmental background noise and half contained competing talkers. The aids were fitted to NAL-NL2 targets (the most-common hearing aid fitting formula) for a standard moderate sloping hearing loss, and Fortell says none of the test audio had been used to train AI 2.0.

Signal-to-noise ratio (SNR) benefit was the primary measurement. Fortell also assessed HASPI v2 and STOI, which estimate speech intelligibility, and PESQ, which evaluates speech quality.

The results indicated significant advantages for AI 2.0 across both AI modes and noise types. Small SNR improvements can make a big difference in some listening situations. As a general rule of thumb, a 1 dB improvement in SNR can translate into roughly an 8-10 percentage-point improvement in speech recognition in noise (although the actual benefit varies with the listener, the hearing loss, and listening conditions). When comparing AI 2.0 versus AI 1.0, here’s what they found:

• In ordinary background noise, SNR benefit increased from 8.9 to 10.6 dB in All Voices and from 9.1 to 10.8 dB in Front Voices—improvements of about 1.7 dB.
• With competing talkers, All Voices increased from 1.9 to 5.9 dB, a 4.1 dB improvement.
• Front Voices, which already performed strongly with competing talkers, reached 11.0 dB of SNR benefit, approximately 1 dB better than AI 1.0.
• Across both modes and noise types, the average SNR improvement was 2.13 dB.
• AI 2.0 produced a better average result in all 16 combinations of mode, noise type, and measurement metric. Fifteen reached statistical significance (the exception was STOI in Front Voices with competing talkers).

Our early take

Fortell is one of the most exciting new hearing aid offerings we have encountered in years. Its central promise is not a modest improvement in a generic noise program, but real-time separation of wanted speech from competing acoustic sources—similar but also different from the strategy used by the current leader in HearAdvisor’s hearing aid rankings for speech in noise, the Phonak Infinio Sphere.

AI 2.0 strengthens the proposition in two ways. First, Fortell’s testing suggests that retraining the model on more real-world audio can improve both speech metrics and sound quality without changing the hardware. Second, every current owner receives the upgrade at no charge, making the hearing aids potentially more capable after purchase. HearingTracker would love to see these types of firmware updates become a continuing trend in hearing aids.

Fortell’s reported 2.13 dB average SNR improvement over AI 1.0 is also impressive. It appears across a broad set of controlled sound scenes, but it remains a manufacturer-reported bench result.

Preliminary reports from HearAdvisor appear to be promising. Once HearAdvisor and HearingTracker complete their testing of Fortell, this page will be updated with results, audio recordings, practical observations, pros and cons, and our final assessment.