Transcribe audio from URLs using Deepgram Nova-3 via MPP micropayments on Tempo L2.
What it does
This MPP endpoint provides access to Deepgram's speech-to-text transcription service through the Locus payment proxy. It accepts a publicly accessible audio URL and returns a transcript using Deepgram's Nova-3 model (with Nova-2, enhanced, and base also available). The endpoint supports multiple languages via BCP-47 codes and offers rich processing options including speaker diarization, smart formatting, punctuation, paragraph splitting, summarization, sentiment analysis, topic detection, intent detection, and named entity extraction.
The endpoint is part of a broader Deepgram service suite hosted on the same MPP gateway that also includes text-to-speech (Aura-2 TTS), text analysis, and model listing endpoints. Payment is per-call via the Tempo method (pathUSD on Tempo L2) at $0.053 per transcription request. The request body requires only a `url` field pointing to the audio file; all other parameters are optional toggles or configuration.
Note that the probe returned 404 on HEAD and GET methods, which is expected since this is a POST-only endpoint as defined in the OpenAPI spec. The OpenAPI schema is well-documented with clear parameter descriptions, though no example response schema is provided. External API reference documentation is available at Deepgram's developer portal, and an LLM-oriented skill file is referenced at paywithlocus.com.
Capabilities
Use cases
- —Transcribing podcast episodes or meeting recordings from a URL
- —Adding speaker labels and paragraphs to interview audio for readability
- —Extracting sentiment, topics, and intents from customer support call recordings
- —Generating summaries of long audio content without listening to the full file
- —Building multilingual transcription pipelines for media content
Fit
Best for
- —Agents needing pay-per-call audio transcription without API key management
- —Transcribing publicly accessible audio files with rich NLP features in one call
- —Developers wanting Deepgram access settled via crypto micropayments
Not for
- —Real-time streaming transcription (this is a batch URL-based endpoint)
- —Transcribing local audio files that are not hosted at a public URL
Quick start
curl -X POST https://deepgram.mpp.paywithlocus.com/deepgram/transcribe \
-H "Content-Type: application/json" \
-d '{
"url": "https://example.com/audio.mp3",
"model": "nova-3",
"smart_format": true,
"diarize": true
}'Example
Request
{
"url": "https://example.com/meeting-recording.mp3",
"model": "nova-3",
"topics": false,
"diarize": true,
"intents": false,
"language": "en",
"punctuate": true,
"sentiment": false,
"summarize": false,
"paragraphs": true,
"smart_format": true,
"detect_entities": false
}Endpoint
Quality
The OpenAPI spec is detailed with clear parameter descriptions and pricing info, but no response schema or example responses are provided. The probe returned 404 because HEAD/GET were used on a POST-only endpoint, so liveness is not confirmed but likely. No crawled documentation was available from the gateway itself.
Warnings
- —Endpoint liveness not directly confirmed — probe used HEAD/GET but endpoint is POST-only; a POST probe would be needed to verify 402 challenge
- —No response schema defined in the OpenAPI spec; response structure must be inferred from Deepgram's own API documentation
- —Currency address 0x20c000000000000000000000b9537d11c60e8b50 assumed to be pathUSD with 6 decimals based on Tempo convention; if decimals differ, the price would change
Citations
- —OpenAPI spec lists Nova-3, Nova-2, enhanced, and base as available STT modelshttps://deepgram.mpp.paywithlocus.com
- —Transcribe endpoint costs 53000 base units via Tempo methodhttps://deepgram.mpp.paywithlocus.com
- —External API reference available at Deepgram developer portalhttps://developers.deepgram.com/reference/deepgram-api-overview
- —LLM-oriented documentation referenced at paywithlocus.comhttps://beta.paywithlocus.com/mpp/deepgram.md