See llms.txt for all machine-readable content.

Back to Templates

Transcribe and summarize Russian audio with Speech Expert forms

Created by

Created by: Speech Expert || speechexpert
Speech Expert

Last update

Last update 7 hours ago

Categories

Quick overview

Upload audio or video through an authenticated n8n form and transcribe it in Russian, English, Kazakh, Kyrgyz, Uzbek, Tajik or Armenian. Speech Expert processes the file asynchronously; the workflow returns a downloadable transcript with an optional summary, or a recoverable error report.

How it works

  1. Receives an authenticated n8n form submission containing one audio/video file, a choice of seven speech languages and an optional short or conversation summary. A blank language selection defaults to Russian.
  2. Validates the upload and selected options, then maps the language to a compatible recognition model using an explicit allowlist. Model selection needs no manual URL editing.
  3. Uploads the binary audio field to the Speech Expert recognizeAsync endpoint with the selected model and language. Recognition uses mono audio without diarization or LLM correction; summary generation is optional.
  4. Polls the operation every five seconds, for up to ten minutes or 120 responses. Paid creation requests are not automatically retried, and a polling timeout does not cancel an existing operation.
  5. Builds a UTF-8 text report from the completed transcription, appending any available summary and the operation ID. If summary generation fails, the successful transcript is preserved. Summaries normally follow the predominant transcript language.
  6. Returns the transcript as a downloadable TXT file. Failures return a JSON report with any recoverable operation ID, so the existing operation can be checked before starting another paid request.

Setup

  1. Create a Speech Expert API key and add an n8n HTTP Header Auth credential (header Authorization with value Api-Key YOUR_KEY) and select it in the Speech Expert HTTP Request steps.
  2. Create an n8n Basic Auth credential and configure it on the form trigger so only authorized users can submit files.
  3. Open the test form, upload a short recording, select its speech language and optionally request a summary. Check the resulting download before publishing the production form. Blank language and summary fields mean Russian recognition without a summary.

Requirements

  • An n8n installation or n8n Cloud account, a Speech Expert account with an active API key and sufficient credits, and one supported audio/video file. Tested with n8n 2.46.6 using built-in nodes. The form default is 200 MiB per file; hosting limits may be lower. Prices depend on the recognition model and optional summary.

Customization

  • Replace the form with an existing workflow input while preserving the audio binary field. To change a recognition model, edit its language entry in Validate input and retain a compatible model/language pair. Review binary storage, execution retention and timeouts for larger files.

Additional info

Reviewer note: please update the generated Russian-only title to "Transcribe audio in seven languages with Speech Expert and optional summaries". The replacement JSON now supports exactly Russian, English, Kazakh, Kyrgyz, Uzbek, Tajik and Armenian.

The selected Kazakh, Kyrgyz, Tajik and Armenian Large models use external audio processing; optional summaries use external text processing. Language selection does not translate the transcript. Setup, model mapping and API details: https://speech-expert.ru/documentation/integrations/n8n/