Hi — we are BRAINIALL, the operator of the hosted API used by this flow.
We published a manual, fail-closed Node-RED flow for Brazilian Portuguese or Spanish recordings. It uses only core nodes and returns diarized words plus SRT and WebVTT captions.
The safeguards are deliberate:
- importing the flow cannot upload anything (rightsAndConsent defaults to false)
- the operator must select one absolute local file and explicitly confirm authorization/consent
- the API key stays in the Node-RED environment
- one request only; there is no automatic retry
- the included Debug node emits metadata only, not the path, key, media, transcript, SRT or VTT
- speaker labels and captions still require human review
Flow Library: PT-BR or Spanish transcription with speaker diarization (flow) - Node-RED
Source, tests and setup: GitHub - fasuizu-br/brainiall-node-red-diarized-transcription-flow: Manual core-only Node-RED flow for authorized PT-BR or Spanish transcription with speaker diarization · GitHub
Release v1.1.0 adds a machine-readable JSON Schema for the flow output and passes 43 local assertions: Release v1.1.0 — machine-readable output schema · fasuizu-br/brainiall-node-red-diarized-transcription-flow · GitHub
The hosted API is external and metered; the repository and flow are public. The JSON Schema describes the flow output contract, not the upstream API response.
Feedback would be especially useful on the Node-RED output shape, failure boundaries and whether a separate reusable subflow would be more useful than the current explicit single-tab example.
