← All website & technology solutions
The Obeinson Voice API turns live and recorded voice into structured, industry-ready data: multi-tenant by design, with HIPAA-conscious redaction for clinics, timeline payloads for studios, and priority routing for dispatch.
Every call follows the same pipeline, whatever the industry. The ingestion layer accepts real-time streams and async batch uploads, the gateway keeps tenants isolated, workers do the heavy lifting, and your systems receive signed JSON.
Real-time voice streams (WebRTC/SIP/WebSocket clients push transcript chunks as they arrive) and a REST endpoint for async file uploads and batch processing.
API-gateway checks — bearer-token verification, per-tenant rate limits and daily session quotas. A tenant's data is never visible to another tenant, in the API or in the webhooks.
Healthcare: PHI redaction, ICD-10/CPT tagging and SOAP formatting. Creative: markers, EDL sequence formatting and timestamping. Dispatch: callsign extraction, priority alert flags and geotagging.
Parsing, entity extraction and payload building run in background workers, so your integration never waits on the AI step. A 5-minute sweep retries anything that slips.
Every payload is signed with HMAC-SHA256. Failures retry with exponential backoff; exhausted retries land in a dead-letter queue you can replay.
HTTP POST to your systems: EHR and billing APIs for healthcare, NLE/DAW automation for creative teams, CAD and emergency dashboards for dispatch.
The same API, shaped by a per-tenant domain profile. Tenants are provisioned with one industry; the parser, entities and payload change with it.
Clinic and telehealth voice notes, redacted before anything else happens.
Studio sessions, interviews and voiceover notes for production teams.
Station call-ins, scanner logs and field traffic for radio operators.
Base URL https://obeinsonenterprise.com/api/voice/v1. Authenticate every call with Authorization: Bearer <your API key>. Keys are issued when your tenant is provisioned — contact us to get set up. This page shows the shape of each endpoint; GET /api/voice/v1 returns the same list from the API itself.
/api/voice/v1/sessionsOpen a streaming session, or submit a batch. A batch fans out into one session per item under a job and is queued immediately.
Request{"kind": "stream", "language": "en-US", "meta": {"patient_visit": "AV-1042"}}
{"kind": "batch", "items": [{"text": "transcript…", "meta": {"project": "EP3"}}]}
Response · 201 / 202{"session": {"id": "vs_A1b2C3d4E5f6", "kind": "stream", "status": "open"}, "webhook": true}
{"job": {"id": "vj_Xy9zAb8Cd7Ef", "status": "queued", "total": 2}, "sessions": ["vs_…", "vs_…"]}
Daily session quota and a per-minute burst limit apply per tenant; the response is 429 when either is hit.
/api/voice/v1/sessions/:id/chunksAppend a transcript chunk while a streaming session is open. Text is merged into the working transcript in order; audio sizes are kept as an audit trail only — raw audio is never uploaded or stored.
Request{"seq": 2, "text": "Patient reports persistent cough…", "speaker": "clinician", "bytes": 18200, "mimeType": "audio/opus"}
Response · 201{"chunk": {"seq": 2, "bytes": 18200}, "session": {"id": "vs_A1b2C3d4E5f6", "status": "open"}}/api/voice/v1/sessions/:id/closeClose the session and queue it for processing: redaction, domain parsing, AI entity extraction and the industry payload. When processing finishes, transcript.completed is dispatched to your webhook.
Response · 202{"session": {"id": "vs_A1b2C3d4E5f6", "status": "queued"}, "webhook": true}/api/voice/v1/sessions/:idSession status. Once completed, the response carries the processed transcript: redacted text, segments with speaker turns, extracted entities and the industry payload.
Response · 200{"session": {"id": "vs_…", "status": "completed", "industry": "healthcare", …,
"transcript": {"text": "…", "segments": […], "entities": […], "payload": {"soap": {…}, "icd10": ["J06.9"], "cpt": ["99213"]}}}}/api/voice/v1/transcripts/:sessionThe processed transcript on its own, without the session envelope. 409 while the session is still open, 200 with "transcript": null while it is queued or processing, 422 if processing failed.
/api/voice/v1/jobs/:idBatch job progress: totals, completed and failed counts, and the state of each session in the job.
/api/voice/v1/sessions/:idDelete a session and everything under it — chunks, transcript and entities — for data-retention compliance. 409 while the session is processing.
/api/voice/v1/usageYour tenant, limits and the last 30 days of usage (requests, chunks, minutes, webhook deliveries).
/api/voice/v1/webhooks/testSends a signed ping event to your registered webhook URL so you can verify signature validation end to end.
Register one HTTPS endpoint per tenant. Every delivery is a standard JSON POST signed with your tenant's secret.
Each delivery carries X-Obeinson-Event, X-Obeinson-Delivery and X-Obeinson-Signature: sha256=<hex>. The signature is the HMAC-SHA256 of the raw request body keyed with your webhook secret (issued with your tenant). Compute it over the exact bytes you received — don't re-serialize the JSON first — and compare with a constant-time equality check.
transcript.completed fires when a session finishes processing and carries the entities and industry payload; ping is the test event. Failed deliveries retry on a 1-minute, 5-minute, 30-minute, 2-hour schedule; after the fifth attempt the delivery is dead-lettered and can be replayed from the admin dashboard once your system is back.
transcript.completed payload (healthcare example){
"event": "transcript.completed",
"sessionId": "vs_A1b2C3d4E5f6",
"industry": "healthcare",
"meta": {"patient_visit": "AV-1042"},
"entities": [{"type": "medication", "value": "lisinopril", "start": 148, "end": 158, "confidence": 0.93}],
"payload": {
"soap": {"subjective": "…", "objective": "…", "assessment": "…", "plan": "…"},
"icd10": ["J06.9"], "cpt": ["99213"],
"redactions": 2
},
"sentAt": "2026-10-09T12:00:00.000Z"
}
Payloads are built and signed per tenant, from per-tenant data, delivered to per-tenant endpoints. There is no cross-tenant query path in the API: every session, transcript, job and delivery lookup is scoped to the authenticated tenant's id. Healthcare tenants additionally have PHI redacted before any AI enrichment step, and audio bytes are never persisted — only sizes and mime types for the audit trail.
Tenants are provisioned by our team with an industry profile, an API key and a webhook secret. Standard covers 200 sessions a day; Pro raises it to 2,000 with priority queueing.
Best forSingle-location clinics, studios and stations.
Best forOrganizations running voice at volume.
Best forNetworks needing their own destination adapters.
Ready to wire voice in? Book a free 15-minute audit and tell us about your integration, or read how we build websites, search and technology solutions.