deepgram-speech
Use for Deepgram speech and voice production workflows: speech-to-text transcription, live captions, diarization, audio intelligence, Aura text-to-speech, Flux and Voice Agent live audio, model selection, cost/limits/privacy checks, artifact custody, and production QA.
Profile →npx skills add calesthio/generative-media-skills --skill deepgram-speech --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
What it does
Use this skill when Deepgram is a candidate provider for transcription, captioning, speech analytics, voice-agent audio, or text-to-speech assets. Treat Deepgram as a production speech platform, not as a generic "make audio" button: choose the endpoint, model, features, and custody plan from the deliverable's latency, accuracy, language, privacy, and artifact requirements.
How it works
The skill documents three main paths for Speech-to-text: pre-recorded REST POST /v1/listen, live streaming WebSocket wss://api.deepgram.com/v1/listen, and Flux /v2/listen for turn-based conversational STT. It specifies model guidance including nova-3 for most production audio, flux-* for conversational agents, and nova-2 as a fallback. It covers transcript formatting options (smart_format, diarization, multichannel, language detection, code-switching), caption generation (SRT/VTT, normalization), and Audio Intelligence features (summarization, topic detection, sentiment, entity detection). For TTS, Aura REST POST /v1/speak and WebSocket, with voice IDs like aura-2-thalia-en and limits such as a 2000-character input limit. For Voice Agent, it describes a WebSocket-based integration including a Settings message and routing between STT, LLM, and TTS, with guidance on when to use Voice Agent versus separate components. It includes workflow recipes for pre-recorded transcription, live captions, Voice Agent, and TTS narration, plus empirical observations and production heuristics about chunking, latency, and diarization.
When to use it
Use Deepgram as the provider when the deliverable requires transcription, captions, speech analytics, voice-agent audio, or TTS assets and you need to select endpoints, models, features, and custody based on latency, accuracy, language, privacy, and artifact requirements. Validate pricing and rate limits on the day of the run due to volatility in pricing and quotas.
What it can touch
The skill declares tools: claude-code, codex, copilot, cursor. It provides guidance to interact with Deepgram APIs (REST and WebSocket endpoints), including model selections like nova-3 and flux-*, features such as smart_format, diarize_model, multichannel, redaction, and various TTS and Voice Agent configurations.
Caveats
It notes that pricing and quotas are volatile and should be verified on the day of the run. It warns that some limits and features (e.g., diarization in streaming, rate limits by plan/region) may vary and require checking the current official docs. It also cautions about performance considerations such as latency and chunking strategies, and instructs to verify model IDs, language support, add-on prices, concurrency, regional availability, and early-access controls before spend or launch.
# Deepgram speech production Use this skill when Deepgram is a candidate provider for transcription, captioning, speech analytics, voice-agent audio, or text-to-speech assets. Treat Deepgram as a production speech platform, not as a generic "make audio" button: choose the endpoint, model, features, and custody plan from the deliverable's latency, accuracy, language, privacy, and artifact requireme
What does the deepgram-speech skill do?
Use for Deepgram speech and voice production workflows: speech-to-text transcription, live captions, diarization, audio intelligence, Aura text-to-speech, Flux and Voice Agent live audio, model selection, cost/limits/privacy checks, artifact custody, and production QA.
How do I install it?
Run `npx skills add calesthio/generative-media-skills --skill deepgram-speech --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From calesthio/generative-media-skills, a repository with 112 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.