Integration boundary
What connects to what
Spatius reference clients demonstrate a layered VAD → OpenAI speech recognition → streaming model → OpenAI TTS → AvatarKit path that can be measured stage by stage.
Prerequisites
- Current OpenAI voice pipeline credentials and access
- Spatius App ID, Avatar ID, and an approved session-token flow
- A defined audio source, format, sample rate, and interruption policy
- A test environment that matches the target browser, device, or server runtime
Implementation path
Build a testable baseline
- 1
Confirm the current OpenAI voice pipeline version and read both linked first-party sources.
- 2
Run the smallest official Spatius sample without the partner dependency to establish an avatar baseline.
- 3
Connect OpenAI voice pipeline at the audio or session boundary described in the architecture section.
- 4
Verify audio format, pacing, interruption, reconnect, cleanup, and credential isolation.
- 5
Record versions, target devices, network conditions, failures, and rollback behavior before production.
Decision guidance
Best for and not best for
Best for
- Teams already committed to OpenAI voice pipeline
- Developers who want to keep the avatar layer separable from the conversation stack
Not best for
- Teams requiring an unverified integration to be treated as production-ready
- Buyers looking for a fully managed end-to-end avatar-video service
Evidence-aware assessment
Advantages and limitations
Advantages
- Adds a OpenAI voice pipeline-specific implementation decision instead of a generic provider list
- Keeps the Spatius rendering boundary explicit
- Provides a concrete validation and failure checklist
Limitations
- Client-side OpenAI keys are for demos only. Move secrets server-side and ensure synthesized speech is converted to mono PCM s16le at the configured sample rate.
- Current package and API versions still require release-time verification
- No first-party benchmark for this exact combined stack is included in the supplied evidence
Client-side OpenAI keys are for demos only. Move secrets server-side and ensure synthesized speech is converted to mono PCM s16le at the configured sample rate. Indexing this guide does not convert an architecture blueprint into a claimed native connector; run and document the validation checklist before production use.
Page-specific FAQ
Implementation questions
Is OpenAI voice pipeline a native Spatius connector?
The supplied evidence classifies this as a documented path. Recheck the linked first-party sources for the current support boundary.
What must be tested before production?
Credentials, audio access and format, sample rate, interruption, reconnect, backpressure, cleanup, privacy, versions, and target-device behavior.
Does this page include performance results?
No. It does not publish latency, FPS, bandwidth, or reliability numbers for this exact combined stack without a reproducible first-party test.
Primary evidence
Sources and verification
Reviewed 2026-08-10. Recheck versions and implementation details before publication.
Continue exploring