Skip to main content

Provider evidence profile · attributable, not endorsed

Deepgram

The Lab's first implemented voice provider, with guarded live paths and labeled deterministic fallbacks.

Deepgram profile detail

This changes presentation only. It never changes access, ownership, provider policy, or what an action can do.

Saved on this device.

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.
  • 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.

ONE Voice Concierge

What would you like to accomplish?

Describe an outcome. ONE will explain up to three existing journeys, then wait for your choice.
Choose the detail for this guidance

This changes presentation only. It never changes access, ownership, provider policy, or what an action can do.

Saved on this device.

Why this view?

Add short explanations where a concept becomes useful. The destination IDs, permissions, costs, and outcome semantics never change with this setting.

Guide only: no action, provider, upload, content save, spending, or scenario run can happen here. Your existing presentation-depth preference may persist.