Inworld STT
Realtime STT-1 · 30 languages · Voice Profile · cloud API
Comparative Scores
Architecture
Evaluate transcription, silence and interruption handling in French independently of LLM hosting. No automatic avatar emotion mapping is validated.
Analysis
Realtime STT-1 is a cloud transcription model with 30 documented languages, Voice Profile signals and controllable turn detection. Speaker diarization is experimental. Inference is a separate LLM service; the Realtime API assembles voice components. RAG and embeddings stay application responsibilities.
Strengths
- WebSocket streaming and file transcription
- 30 documented languages
- Voice Profile signals alongside transcript
- Automatic or client-controlled turns
Weaknesses
- French WER and end-to-end latency unmeasured
- Speaker diarization experimental
- No confirmed self-service on-premise
- Voice Profile is not a validated emotion assessment
STT Capabilities
Bidirectional WebSocket or synchronous file transcription. Automatic turn detection or client-controlled turns. French full-pipeline latency remains unmeasured.
30+ languages
Pricing
On-Demand : 0,15 $/h ; Creator : 0,10 $/h avec 25 $/mois de crédits partagés. STT seul, hors LLM/TTS. Vérifié le 15.10.2026.
| Plan | Subscription/mo | Included | Overage/min | Top-up |
|---|---|---|---|---|
On-Demand (free) Pay-as-you-go above included minutes | Free | 400 min | $0.0025 | Yes |
CreatorRecommended Creator credit balance is shared with TTS/LLM. | $25/mo | 15,000 min | $0.0017 | Yes |
Sovereignty & Compliance
Non confirmé par les sources actuelles / Not confirmed by current sources. Enterprise regions are not proof of self-hosting.
Data residency: Public US; EU/India on separately provisioned Enterprise organizations. Model availability must be confirmed.
Cloud API. No documented self-service on-premise offering; any dedicated deployment requires confirmation.
Inworld STT — Strategic Positioning
Beyond technical specs: where does this tool sit in the ecosystem, what are the risks and strategic implications for GamiWays?
STT-1 is a voice-input candidate; Inference adds a separate LLM hosting option, not validated full-stack performance.
A. Strategic Positioning
Target customer: Developer / Enterprise — gaming, regulated industries, real-time interactive experiences
Realtime STT-1: streaming transcription, 30 languages and Voice Profile signals. EU/India require separately provisioned Enterprise organizations.
B. Competitive Moat
- STT-1 · 30 languages · streaming
- Voice Profile signals and controllable turn detection
Vulnerability: No comparable French WER or full-pipeline latency verified; experimental diarization and contractual deployment limits require testing.
E. Strategic Questions for GamiWays
Sovereignty fit
Public US service. Enterprise EU/India model availability must be confirmed. No documented self-service on-premise.
Build vs. Buy
Evaluate STT separately from Inference LLM hosting; keep the current transcription provider as baseline.
Lock-in risk
Proprietary STT and platform features require an interchangeable application adapter.
Roadmap alignment
French transcription, silence and interruptions must be evaluated on the application workload.
This sheet's verification
Update note: AA-WER public 3,9 % ajouté le 17 octobre 2026. Prix et langues de l’audit du 15 octobre conservés. Latence de bout en bout toujours non publiée.
This date covers provider-specific prices, capabilities and notes. Comparative benchmarks follow the synchronization and methodology shown above.
ELO benchmarks / indices: Artificial Analysis