Starters

Deepgram Starter Apps

Clone a working demo and start building. Every starter is a minimal, runnable app you can extend.

1. Pick Your Feature

What do you want to build?

• Transcribe a file → transcription — send audio/video, get text back (REST, Nova)
• Transcribe a live stream → live-transcription — real-time speech-to-text (WebSocket, Nova)
• Generate speech → text-to-speech — send text, get audio back (REST, Aura)
• Stream speech → live-text-to-speech — real-time text-to-audio (WebSocket, Aura)
• Analyze text → text-intelligence — sentiment, topics, intents, summaries over text you already have (REST, /v1/read)
• Build a voice agent → voice-agent — conversational AI agent (WebSocket, agent.deepgram.com)
• Conversational STT with turn detection → flux — Deepgram Flux STT for voice agents and interactive assistants (WebSocket, /v2/listen)
• Turn-based TTS for a voice agent → flux-tts — Deepgram Flux TTS, streaming synthesis with barge-in (WebSocket, /v2/speak)

There is no audio-intelligence starter. text-intelligence is text-only — it posts text you already have to /v1/read. No {framework}-audio-intelligence repository exists in deepgram-starters for any framework, so don’t construct those URLs. To run intelligence features (summarization, sentiment, topics, intents) over audio, they are query parameters on /v1/listen, not a separate starter: clone the transcription starter for your framework and add the parameters to its existing request. See the api skill for which features /v1/listen supports.

Nova vs Flux STT for speech-to-text: use transcription or live-transcription (Nova, /v1/listen) for general-purpose transcription, captions, and batch workloads. Use flux (Flux STT, /v2/listen) when you need built-in turn detection for conversational audio. See the api skill for a full comparison.

Aura vs Flux TTS for text-to-speech: use text-to-speech or live-text-to-speech (Aura, /v1/speak) for one-shot synthesis, non-English voices, and compressed audio. Use flux-tts (Flux TTS, /v2/speak) when you’re streaming LLM output to a speaker and need a turn lifecycle and barge-in. See the api skill for a full comparison.

Flux TTS starters exist for node, flask, fastapi, django, and java only — these are the five apps Deepgram officially publishes at Flux TTS template apps (opens in a new tab). There is no flux-tts starter for the other frameworks; don’t construct those URLs. For an unsupported framework, start from the api skill’s Flux TTS section and the SDK skills instead.

2. Pick Your Stack

3. Clone and Run

Every starter lives at https://github.com/deepgram-starters/{framework}-{feature} — framework first, feature second. Clone with submodules; each starter vendors two git submodules — its browser frontend at frontend/ and the shared starter contracts at contracts/ — and a plain git clone leaves both directories empty and the app unrunnable:

git clone --recurse-submodules https://github.com/deepgram-starters/{framework}-{feature}.git
cd {framework}-{feature}

Both submodule URLs in .gitmodules are SSH (git@github.com:...) even though both repositories are public, so --recurse-submodules fails with Host key verification failed unless the user has a GitHub SSH key. Without one, rewrite SSH to HTTPS for the clone:

git -c url."https://github.com/".insteadOf="git@github.com:" \
  clone --recurse-submodules https://github.com/deepgram-starters/{framework}-{feature}.git

The starter’s own make init runs git submodule update --init --recursive and installs dependencies, but it inherits the same SSH URLs — it fails identically without a key, so it is the path for users who have SSH set up, not a workaround for users who don’t.

Set your API key and follow the README:

export DEEPGRAM_API_KEY=your_key_here

Get an API key at https://console.deepgram.com (opens in a new tab).

Or scaffold with the CLI

The Deepgram CLI (opens in a new tab) has a scaffolder that finds and clones a starter for you:

dg init --list                       # browse templates
dg init --list --search python       # filter
dg init node-transcription           # clone into ./node-transcription
dg init node-transcription --dir ./my-app

dg init does not solve the submodule problem. It runs a plain clone, so frontend/ and contracts/ land empty, and it still prints Done! … is ready and "status": "success". Adding --install runs the starter’s make check-prereqs && make init, which hits the same SSH URLs and fails with Host key verification failed — and dg init reports success anyway. Without a GitHub SSH key, finish the checkout by hand after dg init:

cd my-app
git -c url."https://github.com/".insteadOf="git@github.com:" \
  submodule update --init --recursive

dg init is also marked alpha, and its templates gallery is a separate list from the matrix below rather than a subset of it. It carries 44 templates with no flux or flux-tts entries; it still lists sinatra-transcription, whose repository is archived; and it lists nextjs-* templates that now redirect out of deepgram-starters to deepgram-devs, which is why there is no nextjs row below. Treat the matrix as authoritative and fall back to git clone. See the cli skill for installing deepctl and for the rest of dg init.

The {feature}-html repos are not starters

The deepgram-starters org also contains transcription-html, live-transcription-html, text-to-speech-html, live-text-to-speech-html, text-intelligence-html, voice-agent-html, flux-html, and flux-tts-html. Do not clone these and do not offer them as starters. Each is the shared browser frontend that a backend starter pulls in as its frontend/ submodule — node-transcription vendors transcription-html, flask-voice-agent vendors voice-agent-html, node-flux-tts and java-flux-tts both vendor flux-tts-html, and so on. Seven of the eight say so in their own README (“This is a frontend submodule - do not use directly”); flux-tts-html carries no such warning but is vendored the same way. None of them serve an API, so none of them run standalone. Clone the backend starter instead and the right frontend arrives with it.

They also invert the naming rule. The starter pattern is {framework}-{feature}, but these are {feature}-html — and the mirror-image names do not exist, so do not construct them: deepgram-starters/html-transcription is a 404. There is no vanilla-HTML row in the matrix because there is no standalone browser starter; for browser-only work, clone the node starter for the feature you want and read its frontend/ directory.

Examples

“I want to build a voice agent in Python” → git clone --recurse-submodules https://github.com/deepgram-starters/fastapi-voice-agent.git

“I need live transcription in my Node app” → git clone --recurse-submodules https://github.com/deepgram-starters/node-live-transcription.git

“I want to add text-to-speech to my Go service” → git clone --recurse-submodules https://github.com/deepgram-starters/go-text-to-speech.git

“I want to analyze audio for sentiment in C#” → git clone --recurse-submodules https://github.com/deepgram-starters/csharp-text-intelligence.git

“I want streaming TTS with barge-in for my Node voice agent” → git clone --recurse-submodules https://github.com/deepgram-starters/node-flux-tts.git

“I want a plain browser/HTML demo” → There is no standalone HTML starter. Clone node-{feature} and work in its frontend/ directory — that is the same browser code the {feature}-html submodule holds.

All Starters

Every URL below is a real, published, non-archived repository, and the table is the complete set: 13 frameworks × 7 features, plus flux-tts for the five frameworks that have it. A cell showing — means that starter does not exist; don’t construct the URL.

Need something more specific?

• Focused feature snippets (one feature, one language, < 50 lines) → recipes skill → https://github.com/deepgram/recipes (opens in a new tab)
• Third-party integrations (Twilio, LiveKit, LangChain, Vercel AI SDK, Discord, etc.) → examples skill → https://github.com/deepgram/examples (opens in a new tab)
• SDK-specific code skills (idiomatic imports, async patterns, gotchas) → npx skills add deepgram/deepgram-{lang}-sdk — see the api skill for the 6 SDKs whose skills are publicly installable.

Related Deepgram skills

• api: consolidated REST + WebSocket API reference
• recipes: minimal runnable feature snippets per language
• examples: full integration examples with third-party platforms
• docs: documentation finder
• cli: deepctl, including dg init for scaffolding a template from the terminal
• setup-mcp: Deepgram MCP server installation