Migrate

Switching STT APIs

A provider-agnostic playbook for switching speech-to-text APIs: put transcription behind an interface, compare on your own audio, and roll out gradually with a fallback.

From Deepgram

Move a Deepgram Nova-3 integration to Speech Revolutions: auth, endpoint mapping, response shapes, and the to_deepgram() escape hatch.

From AssemblyAI

Move an AssemblyAI integration to Speech Revolutions: auth, the upload-submit-poll flow, response fields, and millisecond vs second timestamps.

From OpenAI Whisper API

Move an OpenAI /v1/audio/transcriptions integration to Speech Revolutions: auth, multipart vs presigned upload, the 25 MB limit, and getting word timestamps + diarization in one call.

From ElevenLabs

Move an ElevenLabs Scribe integration to Speech Revolutions: xi-api-key auth, multipart vs presigned upload, the spacing-token quirk, diarization, and timestamps.

From self-hosted Whisper

Move a faster-whisper or whisper.cpp deployment to Speech Revolutions' hosted API: retire the GPU ops, keep word timestamps, and get built-in diarization.