Provider evidence profile · attributable, not endorsed
Deepgram
The Lab's first implemented voice provider, with guarded live paths and labeled deterministic fallbacks.
Available Deepgram tools
These tools are provider-attributable capabilities inside ONE. Their presence does not imply sponsorship, endorsement, or universal superiority.
Read the historical provenance noteStable identifiers, methodology, traces, and sanitized evidence.
ONE Voice Lab grew from an independent Deepgram hiring journey and continued hands-on exploration. It contains no private hiring correspondence, program terms, or non-public roadmap information.
Inspect provider lifecycle statusDiscovery, integration, runtime, health, and benchmark are separate axes.
Feature spotlight · Nova-3 Pharma
Speech recognition for live conversations and recorded audio. English transcription specialized for pharmaceutical vocabulary and medication names. Deepgram documents support for recorded audio and live streams, giving developers a focused option for domain-specific terminology.
Provider-described capability; not a claim of live availability in ONE.
Provider documentation · checked 2026-09-27 ↗- DiscoveryCataloged
- IntegrationContract Tests Passed
- RuntimeDisabled
- HealthDisabled
- BenchmarkIneligible
Inspect the evidence boundaryStable identity, verification, credential readiness, and current limitations.
Evidence boundary
These fields report repository and documentation state. They are not a provider quality ranking or production-readiness claim.
- Stable provider ID
- deepgram
- Evidence
- Repository verified
- Documentation
- No dated provider documentation references are attached to this registry record.
- Last verified
- 2026-08-14
- Credential readiness
- Unconfigured
- Readiness explanation
- A provider-neutral adapter is fixture-validated; live invocation remains disabled until every policy and cost gate is satisfied.
Inspect capability evidenceHuman-readable capability families first; exact contract evidence remains available separately.
Capabilities in ONE
Provider-documented support and installed integration remain attributable and separate.
- Discovery
- Realtime Voice
- Speech To Text
- Text To Speech
Inspect exact capability contractsStable capability IDs, verification, adapters, cost, and benchmark eligibility.
discovery.models
- Support
- Supported
- Verification
- Integration Supported
- Integration path
- Adapter
- Benchmark
- Fixture Only
- Cost-bearing
- No
- Required adapter
- Model Discovery
discovery.voices
- Support
- Supported
- Verification
- Integration Supported
- Integration path
- Adapter
- Benchmark
- Fixture Only
- Cost-bearing
- No
- Required adapter
- Voice Discovery
realtime.conversation
- Support
- Supported
- Verification
- Integration Supported
- Integration path
- None
- Benchmark
- Ineligible
- Cost-bearing
- Yes
- Required adapter
- Realtime Voice
realtime.server-agent
- Support
- Supported
- Verification
- Integration Supported
- Integration path
- None
- Benchmark
- Ineligible
- Cost-bearing
- Yes
- Required adapter
- Realtime Voice
stt.prerecorded
- Support
- Supported
- Verification
- Integration Supported
- Integration path
- Adapter
- Benchmark
- Fixture Only
- Cost-bearing
- Yes
- Required adapter
- Prerecorded Stt
stt.streaming
- Support
- Supported
- Verification
- Integration Supported
- Integration path
- None
- Benchmark
- Ineligible
- Cost-bearing
- Yes
- Required adapter
- Streaming Stt
tts.batch
- Support
- Supported
- Verification
- Integration Supported
- Integration path
- Adapter
- Benchmark
- Fixture Only
- Cost-bearing
- Yes
- Required adapter
- Batch Tts
tts.voice-selection
- Support
- Supported
- Verification
- Integration Supported
- Integration path
- Adapter
- Benchmark
- Fixture Only
- Cost-bearing
- Yes
- Required adapter
- Batch Tts
Limitations
- Live execution still requires server configuration, explicit user action, and the applicable runtime safety gate.
- Repository and mocked evidence do not establish account entitlement, provider availability, or production readiness.
- Core model/voice discovery, prerecorded Speech to Text, and batch Text to Speech are adapter-backed; realtime/session operations remain outside the batch contracts.
Inspect public machine interfacesStable identifiers, methodology, traces, and sanitized evidence.
Public machine record
The API record and MCP tools use this same stable provider ID.
Open official provider sourcesCurated external references, never runtime endpoints.
Official sources
External links are curated metadata sources, not runtime endpoints.