What happened
ClearCaptions provides captioned telephone service for people with hearing loss, and its communications assistants already worked with a form of speech recognition. In September 2020 it asked the FCC to let it caption some calls with automatic speech recognition alone, at first only when no assistant was free. The FCC's National Test Lab, run by MITRE, tested the platform. Its median word error rates of 3.5% to 14.8% compared with 8.9% to 19.5% on average for assisted providers. Its median caption delays were 2.8 to 6.1 seconds, against 5.2 to 17.8 seconds. ClearCaptions reported similar results from its own tests and from a consulting firm that it hired. On 13 April 2021 the FCC's Consumer and Governmental Affairs Bureau granted conditional certification for two years. On 4 April 2023, after reviewing annual reports and the service in operation, the Bureau certified ClearCaptions through 3 April 2028 to caption calls with assistants and speech recognition. That order gives no figures for the automatic captions. The FCC does not name the vendor or show that its recognizer learns from data.
Sources
-
Regulatory document, by authors with no tie to the system.
In recent testing by the National Test Lab, ClearCaptions' median Word Error Rate in various call scenarios ranged from 3.5% to 14.8%, while CA-assisted providers' median Word Error Rate averaged from 8.9% to 19.5%.
-
Regulatory document, by authors with no tie to the system.
Based on our review and investigation of the December 2022 Application, the applications previously filed, ClearCaptions's annual reports, and other relevant information regarding ClearCaptions's actual provision of IP CTS, we find that ClearCaptions will provide IP CTS in compliance with the applicable mandatory minimum TRS standards.
-
Other, by the makers.
ClearCaptions. About us (clearcaptions.com), company timeline. Accessed 1 November 2026.
2021 – Increased call captioning speed and accuracy with Automatic Speech Recognition (ASR).
Details
AI system
- Workflow
-
When no assistant was free, the other party's audio went to a speech-to-text vendor, which returned the captions. ClearCaptions chose its vendor after long testing of how few words the platform inserted, substituted or deleted. ClearCaptions planned to test each vendor's captions manually and algorithmically.
Follow-up
- Conditions for replication
-
Neither ClearCaptions nor its vendors may keep captions, call audio or call content after a call ends. Emergency calls go to an assistant when one is free. The service must keep meeting the FCC's minimum standards, which the FCC may check by monitoring and unannounced visits.
Entry details
- Sector
-
Telecommunications relay services / accessibility for people with hearing loss.
- Purpose
-
ClearCaptions built its fully automatic mode, on a speech-to-text vendor's platform, to caption calls for people with hearing loss when no communications assistant is free.
30 more questions not yet checked
model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.
Cite this record
Peter H. Vartanian. (2026). ClearCaptions' automatic captions beat assisted ones in FCC lab tests (AIBD-1000). Artificial Intelligence Benefits Database, edition of 1 November 2026. https://petervartanian.xyz/aibd/records/AIBD-1000.html (archived at https://doi.org/10.7910/DVN/AADSII, version 7.0)