InnoCaption's automatic captions top human-assisted ones in FCC tests

In the FCC's National Test Lab, InnoCaption's automatic captions had average error rates of 4.2% to 12.0%, against 8.0% to 19.7% for human-assisted providers. They also arrived seconds sooner.

  • 1 a person

    Beta users could choose captions with no person listening, and they gave four or five stars to 84% of more than 11,000 automatic calls.

Depth runs from 1, a convenience or time saved, to 5, a life or a livelihood preserved.

Status
Reported benefit.
Evidence

foundstudiedreportedclaimed

An independent regulatory document supports the benefit.

Compared with

Average error rates of CA-assisted providers other than InnoCaption, on the same call scenarios. Measured.

Without the system
8.0% to 19.7%
With it
4.2% to 12.0%
The system’s part
The National Test Lab measured the automatic captions against those of human-assisted providers on the same call scenarios, which isolates the recognizer's output. The beta ratings have no such comparison on the same calls. Google describes Speech-to-Text as offering trained models, including one for phone calls.
Limits
Not yet checked.
Costs and harms
Not yet checked.
Later
Not yet checked.

Download data Cite Suggest a correction

Benefit type
Improved Task Accuracy
System
InnoCaption ASR Calling Feature (fully automatic IP CTS on Google Cloud Speech-to-Text)
Made by
Mezmo Corporation d/b/a InnoCaption, with Google Cloud Speech-to-Text
Used by
Mezmo Corporation d/b/a InnoCaption
Where
United States
Stage
In a pilot
Date
, when the source was published
Kind
Aggregate result
Broad benefit
Documenting/Recording Information
Collections
Scrutiny and findings
Last revised
31 October 2026

Sources

  1. Regulatory document, by authors with no tie to the system.

    Federal Communications Commission, Consumer and Governmental Affairs Bureau. Memorandum Opinion and Order, DA 20-1476, CG Docket Nos. 13-24 and 03-123 (conditional certification of Mezmo Corporation d/b/a InnoCaption to provide fully automatic IP CTS). Released 11 December 2020.

    Indeed, InnoCaption’s ASR feature outperformed the average accuracy rates of CA-assisted services on all five call scenarios included in the National Test Lab testing.

  2. Other, by the makers.

    InnoCaption. Home page (innocaption.com). Accessed 31 October 2026.

    Depending on your needs, you can choose between two different caption modes: best-in-class AI speech recognition and live human stenographers, with the ability to toggle between them mid-call at the press of a button.

  3. Other, by the makers. Context or contrary evidence.

    Google Cloud. Speech-to-Text product page. Accessed 31 October 2026.

    Choose from a selection of trained models for voice control and phone call and video transcription optimized for domain-specific quality requirements.

Details

AI system

Workflow

Captions came from the Google Cloud Speech-to-Text API with InnoCaption's own software. InnoCaption tested several engines for speech recognition before choosing Google's. Users could switch between automatic and assisted captions during a call.

Follow-up

Conditions for replication

Call audio and transcripts are not stored or used for machine learning. Calls to 911 go to a communications assistant.

Entry details

Sector

Telecommunications relay services / accessibility for people with hearing loss.

Purpose

Mezmo built InnoCaption's automatic mode to give people with hearing loss captions of their phone calls without a communications assistant.

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). InnoCaption's automatic captions top human-assisted ones in FCC tests (AIBD-0823). Artificial Intelligence Benefits Database, edition of 31 October 2026. https://petervartanian.xyz/aibd/records/AIBD-0823.html (archived at https://doi.org/10.7910/DVN/AADSII, version 4.0)

BibTeX Data Field definitions