Using Deepgram Speech-to-Text (Rust SDK)

Use this skill for prerecorded transcription, live streaming transcription, or when mapping Deepgram docs to the Rust crate's real listen surface.

When to use this product

• Transcribing local files, URLs, or in-memory audio with Deepgram::transcription().
• Streaming audio over WebSocket with stream_request() / stream_request_with_options(...).
• Using common::options::Options for STT features such as model, language, punctuate, diarize, smart_format, utterances, and streaming knobs like endpointing.

This skill covers Nova models on /v1/listen — Deepgram's general-purpose STT family (nova-3, nova-2, nova, enhanced, base). Both Nova and Flux are actively maintained, industry-leading STT model families.

Use a different skill when:

• You need conversational-audio transcription with built-in turn detection (voice agents, interactive assistants) → deepgram-rust-conversational-stt (Flux on /v2/listen).
• You want analytics overlays on the transcript (summarize, sentiment, topics, intents) → deepgram-rust-audio-intelligence (same /v1/listen endpoint, different params).
• You need a full-duplex voice agent (STT + LLM + TTS in one WSS) → deepgram-rust-voice-agent.

Authentication