Kintsugi
Voice biomarker AI that flagged depression and anxiety from 20 seconds of speech; commercial ops retired, models open-sourced.
Retired
Kintsugi appears to be discontinued or shut down — Kintsugi shut down commercial operations in early 2026 after failing to secure FDA clearance and released its voice biomarker models as open source. This page is kept as an archive record and is excluded from search engines.
About
Kintsugi built machine learning models that detect clinical markers of depression and anxiety from just 20 seconds of free-form speech, analyzing how someone speaks rather than what they say. Its KiVA voice biomarker API was aimed at call centers, telehealth visits, and remote patient monitoring, and the company claimed over 80 percent clinical accuracy across multiple languages. A peer-reviewed cross-sectional study in Annals of Family Medicine (2025) tested the tool on nearly 15,000 participants and reported it identified depression in about 71 percent of cases and ruled it out in close to 74 percent.
Kintsugi shut down its commercial operations in early 2026 after failing to secure FDA clearance in time to sustain the venture-backed model, and released its voice biomarker models and scientific methodologies as open source instead. The company had raised about $28 million across a 2021 seed and a 2022 Series A from investors including Insight Partners, with NSF grant support.
What it can do
- Voice biomarker depression detection
- Anxiety screening from 20 seconds of speech
- Language-agnostic acoustic analysis
- Call center and telehealth API integration
- Peer-reviewed validation study
- Open-sourced models after shutdown
Setup
Commercial product discontinued; open-sourced models and methodologies are the remaining artifact.
Prerequisites
None for research use of the open-sourced materials.
Use cases
Works with
Trust score — 5/10
Scored 5/10 (TIER metrics, metrics-v1). Evidence: Healthcare IT News confirms the early-2026 shutdown after the failed FDA bid and the open-source release; Healio confirms the peer-reviewed Annals of Family Medicine study (71 percent sensitivity, 74 percent specificity); $28M in funding confirmed. Strengths: peer-reviewed validation, genuinely novel biomarker approach, open-sourced legacy. Docked for: (1) commercial operations retired, no supported product remains; (2) pre-shutdown accuracy claims were vendor-reported. License: models open-sourced at shutdown; commercial terms discontinued.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 4.0/10
- Maintenance · recency, releases, issue load
- 2.0/10
- Openness · license + docs
- 8.0/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · Commercial operations shut down early 2026 after failed FDA clearance bid
- · Peer-reviewed Annals of Family Medicine 2025 study: ~71% sensitivity, ~74% specificity on 14,898 participants
- · Raised ~$28M (2021 seed $8M, 2022 Series A $20M) plus NSF grants
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.