Speech-to-Text API

The speech-to-text API that keeps your audio in Germany

A REST API with 20 endpoints, 99 languages, speaker diarization on every tier and HMAC-signed webhooks. Our own engines on our own hardware in Germany, no third party in the speech pipeline, from €0.18 per hour of audio.

Base URL: https://api.deepscript.com/v1
Authentication: Bearer ds_live_…
Rate limit: 100 requests/min authenticated, 30 without a key
Data residency: Germany

Three calls to a transcript

The upload responds immediately with an ID. From there, poll or wait for the webhook.

# 1. Upload
curl -X POST https://api.deepscript.com/v1/transcriptions \
  -H "Authorization: Bearer ds_live_YOUR_KEY" \
  -F "file=@meeting.mp3" -F "model=standard" -F "language=auto"

# 2. Poll
curl https://api.deepscript.com/v1/transcriptions/{id}/status \
  -H "Authorization: Bearer ds_live_YOUR_KEY"

# 3. Export
curl "https://api.deepscript.com/v1/transcriptions/{id}/export?format=srt" \
  -H "Authorization: Bearer ds_live_YOUR_KEY" -o result.srt

Error format: RFC 7807 Problem Details, application/problem+json.

Every endpoint

Paths, methods and names are English, exactly as in the OpenAPI specification. Try them interactively with your own key in the reference.

Transcriptions

The core of the API: upload a file, poll for status, fetch the result in four formats. The upload responds immediately with an ID, processing runs asynchronously.

Method	Path	Summary
POST	/v1/transcriptions	Create transcription (file upload)
GET	/v1/transcriptions	List transcriptions
GET	/v1/transcriptions/{id}	Get transcription detail
DELETE	/v1/transcriptions/{id}	Delete transcription
GET	/v1/transcriptions/{id}/status	Get transcription status (lightweight polling)
GET	/v1/transcriptions/{id}/events	SSE stream for real-time status updates
GET	/v1/transcriptions/{id}/export	Export transcription

Vocabularies

Custom word lists for product names, domain terms and proper nouns. Created once, referenced by ID on any transcription.

Method	Path	Summary
GET	/v1/vocabularies	List vocabularies
POST	/v1/vocabularies	Create vocabulary
GET	/v1/vocabularies/{id}	Get vocabulary
PUT	/v1/vocabularies/{id}	Update vocabulary
DELETE	/v1/vocabularies/{id}	Delete vocabulary

Webhooks

Instead of polling: register an HTTPS endpoint and receive an HMAC-SHA256 signed delivery on every event, with three retry attempts.

Method	Path	Summary
GET	/v1/webhooks	List webhook endpoints
POST	/v1/webhooks	Create webhook endpoint
DELETE	/v1/webhooks/{id}	Delete webhook endpoint

Account

Workspace balance, usage and profile. Useful to check whether the balance covers a large batch before starting it.

Method	Path	Summary
GET	/v1/account/balance	Get current balance
GET	/v1/account/usage	Get usage analytics
GET	/v1/account/profile	Get account profile

System

Callable without authentication. The language list is the authoritative answer to which languages are actually supported.

Method	Path	Summary
GET	/v1/health	Health check
GET	/v1/languages	List supported languages

Webhook events

Every delivery carries an HMAC-SHA256 signature in the header and is retried three times with increasing backoff on failure.

transcription.completedSent when a transcription completes
transcription.failedSent when a transcription fails
balance.lowSent when workspace balance drops below the configured threshold

Machine readable

The complete OpenAPI 3.1 specification is mirrored on this domain. Client generators, Postman and AI agents can read it directly.

/openapi.json Interactive reference with try-it-out

Wire it up without code

The same API in different wrapping: MCP server for AI agents, no-code via Zapier or Make, self-hosted via n8n.

REST API for transcription – three calls, doneUpload, poll, fetch. JSON, word-level timestamps, 99 languages, own servers in Germany.MCP Server – AI agents read your transcripts directlyClaude Desktop, Cursor, Zed and other MCP clients talk to your DeepScript library directly. No copy-paste, no third-party cloud.Webhooks – DeepScript calls you when the job is doneNo polling, no long-polling, no SSE waiting game. HMAC-signed POSTs with a retry policy.Zapier Integration – transcription without a single line of codeDrop audio in Google Drive, get a Slack ping when done, create a Notion page with the transcript – built in 3 minutes.Make.com Integration – audio pipelines with branching and iterationDrive watcher, transcription, GPT summary, email digest – all in one visual scenario.n8n Integration – self-hosted audio automationYour n8n + your DeepScript servers = full data sovereignty. No cloud automation, no Schrems II exposure.

API key in two minutes

Create an account, generate a key, start calling. The first 3 transcriptions are free, no credit card.

Start for free See pricing