Description
n8n-nodes-nexara
This is an n8n community node. It lets you use the Nexara
speech-to-text API in your n8n workflows.
Nexara transcribes and diarizes audio, tags speaker roles, scores segments for
emotion, and can run an LLM over the transcript to produce structured output. Full API documentation lives
at docs.nexara.ru.
n8n is a fair-code licensed
workflow automation platform.
Installation
Operations
Credentials
Usage
Resources
Installation
Follow the
installation guide
in the n8n community nodes documentation.
In n8n: Settings → Community Nodes → Install, then enter
n8n-nodes-nexara.
Operations
- Transcribe — transcribe audio to text and wait for the result.
- Diarize — transcribe and label speakers, with optional speaker roles.
- Create Job — submit audio for deferred (long-audio) processing. Returns a
- Get Job — fetch a deferred job by its ID.
- Response Format — JSON, Verbose JSON, Text, SRT, or VTT.
- Language — ISO-639-1 code (e.g.
ru,en), or empty for auto-detection. - LLM Prompt / LLM JSON Schema — run an LLM over the transcript and,
- Timestamp Granularity, Profanity Filter, Dictionary, Model.
- Diarization: Number of Speakers, Diarization Setting, Roles
- n8n community nodes documentation
- Nexara API documentation
job immediately, or optionally polls until it finishes.
Each operation accepts audio either as a binary file from a previous node
or as a URL for the server to download.
Options
optionally, force structured JSON output. A prompt forces Verbose JSON.
(auto, a JSON array like ["client","agent"], or a JSON object mapping
label to description), Emotions.
Credentials
You need a Nexara API key. Create one in the Nexara dashboard, then in n8n
create a Nexara API credential and paste the key. The key is sent as a
Bearer token on every request.
> Note: the Nexara API currently exposes no lightweight health endpoint, so
> the credential has no automated “Test” button. Run a Transcribe operation on a
> short clip to confirm the key works.
Usage
Transcribe an uploaded file
1. A node that produces binary audio (e.g. HTTP Request, Read Binary File).
2. Nexara → Transcribe, Input Type Binary File, Input Binary Field data.
Diarize a call with named roles
1. Nexara → Diarize.
2. Under Diarization Options, set Roles to ["client","agent"].
Score a call for emotion
1. Nexara → Diarize, and under Options set Model to nexara-ru.
2. Under Diarization Options, turn on Emotions.
Each segment that could be scored comes back with an emotion object — label
(angry, sad, neutral, or positive), confidence, and a probs
distribution. Segments the model could not score simply have no emotion key,
so branch on its presence rather than assuming it. Emotion is produced by the
nexara-ru model itself, so it needs that model, the Diarize task, and a JSON
or Verbose JSON response format; the node rejects the other combinations before
sending, since the server bills the surcharge per second and would answer 400
anyway.
Long audio (deferred)
1. Nexara → Create Job with Wait for Completion enabled — the node polls
until the job finishes and returns the result. A failed job is never billed.
2. Or leave polling off and use Get Job later (from another workflow or run)
with the returned Job ID. Results live for 12 hours after creation.
