Skip to main content
Medera ships a BH-tuned speech engine (Speech-to-Behavioral-Text) and a language-agnostic vocal-acoustic engine. Language support varies by surface — the matrices below show what’s GA, what’s in private beta, and what’s on the roadmap.
The vocal-acoustic engine — F0, jitter, shimmer, HNR, prosodic flatness, depression / anxiety / distress indices — is language-agnostic and ships unchanged across every locale. The same is true for the facial physiological engine and the neurobehavioral construct computer.

​
Speech-to-Behavioral-Text — per locale

​
Surface coverage matrix

​
Translation policy

Patient-facing translations preserve clinical fidelity. Validated screeners use the published validated translation (Spitzer et al. for PHQ-9, Sawaya et al. for the Arabic validation) — never a model-generated one.

​
Output language

Even when the input language is non-English, structured emission stays in clinical English — DSM-5-TR labels, ICD-10 codes, C-SSRS bands, instrument names — so the data slots directly into your EHR. This is configurable: set output_language: "source" to emit labels in the source language.

​
What’s next

https://mintcdn.com/medera-357fd587/iC-6rAuwl-0aQSw_/icons/ui/audio-lines.svg?fit=max&auto=format&n=iC-6rAuwl-0aQSw_&q=85&s=a4f3136aa8df6fe2d044778c71ba2bbc

Speech-to-Behavioral-Text

Three endpoints, one BH-tuned engine.
https://mintcdn.com/medera-357fd587/iC-6rAuwl-0aQSw_/icons/ui/globe.svg?fit=max&auto=format&n=iC-6rAuwl-0aQSw_&q=85&s=c6381a12d6805e01a35cb84787415b5b

Data residency

US / EU / MENA / customer VPC.
https://mintcdn.com/medera-357fd587/iC-6rAuwl-0aQSw_/icons/ui/link.svg?fit=max&auto=format&n=iC-6rAuwl-0aQSw_&q=85&s=44f5f1ae1b0122daa0138e9f47b49449

Sub-processors

Per-region sub-processor list.
https://mintcdn.com/medera-357fd587/n1mgcc3nR2sq_PoL/icons/ui/shield-check.svg?fit=max&auto=format&n=n1mgcc3nR2sq_PoL&q=85&s=6909d120dcab3cdfef838294f042d00c

Compliance overview

HIPAA + 42 CFR Part 2.