Founded in 2014 as Transcense by a Deaf-hearing founding team, Ava captions live group conversations for deaf and hard-of-hearing users, tracking multiple speakers across devices. Its AI reaches about 90% accuracy alone and up to 99% with human-assisted Ava Scribe, and is cited b
~90 %
Fully automated AI captioning accuracy (2014-present)
accessibility technology, AI speech-to-text, workplace and education inclusion
Context
Ava (originally Transcense) was founded in San Francisco in 2014 by Thibault Duchemin, Skinner Cheng and Pieter Doevendans, growing out of UC Berkeley's Big Ideas student contest and combining Deaf and hearing founders' perspectives on live captioning.
Objectives
The product aims to caption group conversations in real time by having each participant connect their own phone, so multiple speakers can be tracked and captioned simultaneously — a harder problem than the single-speaker dictation most captioning tools handle.
Activities
Ava integrates with Zoom, Google Meet and Webex and is used in workplaces, classrooms, healthcare and retail settings; its human-assisted Ava Scribe tier pairs AI with live captioners for high-stakes settings such as legal or medical appointments.
Results
The company reports roughly 90% accuracy for its fully automated AI captioning and up to 99% accuracy with human-assisted Ava Scribe. The European Commission's AccessibleEU centre lists 'Ava Captions for All' as a Belgian digital-accessibility case study, and the company has raised over $16 million in venture funding, including a $10 million Series A led by Khosla Ventures and Initialized Capital.
Conclusions
These figures come from company reporting, press coverage and an EU case-study listing rather than an independent, peer-reviewed accuracy audit, so they should be read as corroborated but self-reported performance claims.
Implementation
Indicative cost
Medium (€50k–€500k)
Time to results
Long (> 3 years)
Staffing & skills
Deaf-hearing founding team (Thibault Duchemin, Skinner Cheng, Pieter Doevendans), Human live captioners for the Ava Scribe human-assisted tier in high-stakes settings
Conditions for success
Integration with mainstream video-conferencing platforms (Zoom, Google Meet, Webex) for adoption in existing workflows
Sustained venture funding (over $16M raised) to fund continued AI development
A human-assisted tier for settings where near-perfect accuracy is essential (legal, medical)
Where it fits
Scale
individual/organisational (SaaS product, not a government programme)
Do you run this practice?
Claim it —
verified implementers get a public contact pathway and can propose corrections.
Data sources
Where this practice's information was retrieved from, and when.
A Vietnamese NGO where blind and low-vision engineers lead development of free screen-reader-compatible software (Braille transcription, text-to-speech, e-book reading); Zero …
