Fortell has announced AI 2.0, a software update to its AI-powered hearing aids that the company says improves speech clarity in challenging noisy and multi-talker environments.
The update builds on the AI processing platform introduced with Fortell’s hearing aids in December 2025. Rather than introducing new hardware, AI 2.0 uses an expanded set of real-world training data along with updates to several underlying signal-processing algorithms.
According to Fortell, its team collected training audio in environments where hearing aid users commonly experience difficulty, including restaurants, cafés, construction sites, trains, buses, cars, reverberant offices, and living spaces. The company’s neural network was then retrained using the expanded audio dataset.
Testing AI 2.0 Against the Original Model
To evaluate the update, Fortell compared AI 2.0 with its original AI model, now referred to as AI 1.0, using identical acoustic scenes and the same pair of Fortell AI Hearing Aids.
The testing used a KEMAR acoustic manikin wearing the hearing aids and included 80 challenging speech-in-noise scenes. Forty incorporated background noise from real-world environments such as cafés and train stations, while another 40 included three competing talkers positioned around the listener.
Both AI versions were evaluated using identical hearing aid settings in Fortell’s two AI listening programs:
- All Voices: Designed to reduce background noise while keeping voices from multiple directions audible.
- Front Voices: Designed to emphasize speech coming from directly in front of the listener while suppressing competing talkers from other directions.
The researchers evaluated performance using four objective measures: signal-to-noise ratio (SNR), Hearing-Aid Speech Perception Index version 2 (HASPI v2), Perceptual Evaluation of Speech Quality (PESQ), and Short-Time Objective Intelligibility (STOI).
Fortell Reports Improvements Across All 16 Comparisons
According to the company’s analysis, AI 2.0 produced an improvement over AI 1.0 in the mean result for all 16 combinations of listening program, acoustic condition, and objective metric. Fifteen of the 16 comparisons reached statistical significance.
Across both listening modes and noise conditions, Fortell reported an average 2.13 dB improvement in SNR benefit with AI 2.0 compared with AI 1.0.
The largest change occurred in All Voices mode during competing-talker conditions. In that test, SNR benefit increased from 1.9 dB with AI 1.0 to 5.9 dB with AI 2.0, representing a 4.1 dB improvement.
Figure 1. Background-noise scenes comparing AI 1.0 and AI 2.0 in All Voices and Front Voices modes across SNR, HASPI v2, PESQ, and STOI measures.
In background-noise conditions, output SNR benefit increased from 8.9 to 10.6 dB in All Voices mode and from 9.1 to 10.8 dB in Front Voices mode. These represented per-scene improvements of 1.6 dB and 1.7 dB, respectively. Fortell also reported improvements in PESQ, HASPI v2, and STOI in both programs.
Figure 2. Competing-talker scenes comparing AI 1.0 and AI 2.0 in All Voices and Front Voices modes across SNR, HASPI v2, PESQ, and STOI measures.
In competing-talker conditions, Front Voices achieved an 11.0 dB output SNR benefit with AI 2.0, a 1.0 dB improvement over AI 1.0. All Voices showed the larger change, increasing from a 1.9 dB SNR benefit with AI 1.0 to 5.9 dB with AI 2.0.
HASPI v2, PESQ, and STOI also improved in the All Voices competing-talker condition. The only one of the 16 comparisons that did not reach statistical significance was STOI for Front Voices in competing-talker scenes.
A Software-Based Update
Fortell says the performance gains come from retraining its neural network on the expanded real-world audio dataset and making additional improvements to its signal-processing algorithms. Because the testing used the same physical hearing aids for both AI versions, the reported differences were not associated with changes to the hardware.
AI 2.0 is being delivered as a free over-the-air software update to existing Fortell AI Hearing Aids, making the updated processing available to current wearers without requiring new devices.
The findings are based on objective laboratory measures rather than a clinical trial comparing wearer outcomes with AI 1.0 and AI 2.0. Fortell has published its methodology, statistical analysis, and complete results in an accompanying technical white paper.
About Fortell
Fortell develops AI-powered hearing technology designed to improve clarity in conversation. The company brings together scientists, engineers, audiologists, designers, and other specialists focused on advancing hearing technology through artificial intelligence, custom hardware, and human-centered hearing care. Fortell AI Hearing Aids are prescription devices for adults with hearing loss and are currently available in the United States.
References:
Bisgaard N, Vlaming MSMG, Dahlquist M. (2010). Standard audiograms for the IEC 60118-15 measurement procedure. Trends Amplif. 2010;14(2):113–120. doi:10.1177/1084713810379609
Chung JS, Nagrani A, Zisserman A. (2018). VoxCeleb2: Deep speaker recognition. Proc. Interspeech 2018, 1086–1090. doi:10.21437/Interspeech.2018-1929
Fortell Research. (2026). Spatial AI improves speech intelligibility for hearing aid wearers in challenging multi-talker noise. Fortell Research white paper, Vol. 1. https://www.fortell.com/evidence
Hagerman B, Olofsson Å. (2004). A method to measure the effect of noise reduction algorithms using simultaneous speech and noise. Acta Acust. united Acust. 2004;90(2):356–361.
Kates JM, Arehart KH. (2021). The Hearing-Aid Speech Perception Index (HASPI) version 2. Speech Commun. 2021;131:35–46. doi:10.1016/j.specom.2020.05.001
Morris C, Lovchinsky I, Callahan C, Wallace K, Berry M, Lyngaas K, Agmon N, Malkin I, Nagler B, Montano J, Weinstein B. (2026). Spatial AI consistently preferred to state-of-the-art hearing aids in multitalker noise. Int. J. Audiol. 2026:1–13. doi:10.1080/14992027.2026.2663345
Rix AW, Beerends JG, Hollier MP, Hekstra AP. (2001). Perceptual evaluation of speech quality (PESQ)—a new method for speech quality assessment of telephone networks and codecs. Proc. IEEE ICASSP 2001, 749–752. doi:10.1109/ICASSP.2001.941023
Taal CH, Hendriks RC, Heusdens R, Jensen J. (2011). An algorithm for intelligibility prediction of time-frequency weighted noisy speech. IEEE Trans. Audio Speech Lang. Process. 2011;19(7):2125–2136. doi:10.1109/TASL.2011.2114881
Weisser A, Buchholz JM, Oreinos C, Badajoz-Davila J, Galloway J, Beechey T, Keidser G. (2019). The Ambisonic Recordings of Typical Environments (ARTE) database. Acta Acust. united Acust. 2019;105(4):695–713. doi:10.3813/AAA.919349
Source: Fortell
