{Tech: Europe} London AI Hackathon

Welcome to the {Tech: Europe} London AI Hackathon! We brought voice AI APIs and Lego boxes. You bring the ideas. The SLNG team is here all weekend. Come find us.

Look for Ismael Ordaz (founder) and Nicolas Grenié (DevRel). We’re happy to help you build.

What is SLNG?

SLNG is one platform for real-time speech. Text-to-Speech, Speech-to-Text, and Voice Agents, all under 200ms. We aggregate providers like Deepgram, Rime, Cartesia, and Kugel behind a single endpoint. You switch models by changing the URL, not your code. Underneath, an execution layer caches and routes each turn to cut latency and cost. See the execution layer.

Text-to-Speech

Text in, audio out. Multiple voices and languages.

Speech-to-Text

Transcribe audio files or live streams.

Voice Agents

Full voice agents that make and receive phone calls.

Fastest start: let your coding agent build it

Using Claude Code or another coding agent this weekend? Install the SLNG skills pack and your agent can set up your key, synthesize speech, transcribe audio, and build voice agents on its own.

npx skills add slng-ai/skills

Then ask your agent to build. Prefer to wire it up yourself? Do it the old way below.

SLNG agent skills

See what each skill teaches your agent to do.

Building in n8n?

We just shipped an n8n community node for this hackathon. It drops SLNG into any workflow: generate speech, transcribe audio, and dispatch outbound voice-agent calls. A trigger node also lets a voice agent call your workflow as a tool mid-call. Install it from your n8n instance under Settings → Community Nodes using the package name n8n-nodes-slng. See the n8n node page for the actions and trigger, or the GitHub repo for workflow templates.

The old way, write code

Create an account

Generate an SLNG key

Go to the API Keys page in your dashboard and create a new key. Copy it somewhere safe. You won’t be able to see it again.Every SLNG_API_KEY placeholder in the snippets below is the key you just copied.

Test it

Run this in your terminal to make sure everything works:

curl https://api.slng.ai/v1/bridges/unmute/tts/slng/deepgram/aura:2-en \
  -H "Authorization: Bearer SLNG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "Hello from London!"}' \
  --output hello.wav

You should get a hello.wav audio file. If you hear “Hello from London!”, you’re all set.

Start building

Text-to-Speech
Speech-to-Text
Voice Agents
Unified API

Send text, get audio back.

# Change this to swap models (e.g. slng/rime/arcana:3-en, slng/rime/arcana:3-es)
MODEL="slng/deepgram/aura:2-en"

curl "https://api.slng.ai/v1/bridges/unmute/tts/${MODEL}" \
  -H "Authorization: Bearer SLNG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Hello from London!"
  }' \
  --output hello.wav

Pick a different voice by adding "voice": "aura-2-theia-en" to the request body. See all available voices.

Other languages

Switch the model in the URL. The rest of the code is identical. Spanish via Rime Arcana v3:

curl https://api.slng.ai/v1/bridges/unmute/tts/slng/rime/arcana:3-es \
  -H "Authorization: Bearer SLNG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "¡Hola desde Londres!"}' \
  --output hola.wav

Spanish, Hindi, English (HTTP): slng/rime/arcana:3-es, slng/rime/arcana:3-hi, slng/rime/arcana:3-en
French (WebSocket only): Cartesia Sonic 3 or Kugel

See the Models page for the full language matrix.

Upload an audio file and get a transcription back.

# Change this to swap models (e.g. slng/deepgram/nova:3-multi for multilingual, soniox/soniox:latest)
MODEL="slng/deepgram/nova:3-en"

curl "https://api.slng.ai/v1/bridges/unmute/stt/${MODEL}" \
  -H "Authorization: Bearer SLNG_API_KEY" \
  -F "audio=@recording.wav"

Transcribing French, Spanish, or mixed-language audio? Switch the model to slng/deepgram/nova:3-multi. Same code, auto-detects across 10+ languages.Need real-time transcription? Check out the WebSocket STT examples.

Voice Agents pair an LLM with STT and TTS so you get a full conversational agent that can take and make phone calls or run in a browser. Create one through the Agent Studio UI or via the API.You pick the models, write a system prompt, and optionally add tools (webhooks to your backend, hangup, human transfer, date/time lookup). The agent handles the rest.

curl -X POST https://api.agents.slng.ai/v1/agents \
  -H "Authorization: Bearer SLNG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Hackathon Demo Agent",
    "system_prompt": "You are a helpful assistant at a hackathon. Keep responses short, one or two sentences. Be friendly and direct.",
    "greeting": "Hey! I am your hackathon assistant. What can I help you with?",
    "language": "en",
    "models": {
      "stt": "slng/deepgram/nova:3-en",
      "llm": "groq/openai/gpt-oss-120b",
      "tts": "slng/deepgram/aura:2-en",
      "tts_voice": "aura-2-thalia-en"
    },
    "tools": [
      {
        "type": "template",
        "id": "hangup-tool",
        "template": "hangup",
        "execution_policy": {
          "pre_action_message": { "enabled": true, "text": "Good luck with your project!" }
        }
      }
    ],
    "region": "eu-central",
    "slng_api_key": "SLNG_API_KEY"
  }'

You can also skip the API and build your agent visually in Agent Studio, then test it right in the browser.Want to go further? Add webhook tools to connect your agent to external APIs, or use template variables to make prompts dynamic. See the full Voice Agents guide.

All the TTS and STT examples above go through the Unified API. One endpoint covers every provider, and you swap a model by changing the URL. Handy when you want to try multiple providers without writing a different integration for each one.

https://api.slng.ai/v1/bridges/unmute/{stt|tts}/{provider}/{model}:{variant}

The bridges/unmute/ path segment is the legacy route name. Same endpoint, same Unified API. Works over both HTTP and WebSocket on the same URL.Need fine-grained control over a specific provider’s features? Call each model directly.See the full Unified API documentation.

Add it to your existing agent

Already built on LiveKit or Pipecat? Swap in SLNG by changing one model string. Same pipeline, any provider behind one key.

https://mintcdn.com/slng/LujG4G8gXFrAV3XP/images/providers/livekit.svg?fit=max&auto=format&n=LujG4G8gXFrAV3XP&q=85&s=5ae1614bbad360d24b36276e04bbc0ea

LiveKit plugin

livekit-plugins-slng adds STT and TTS adapters for LiveKit Agents.

https://mintcdn.com/slng/HUXme6ATYhzzt7mD/images/providers/pipecat.svg?fit=max&auto=format&n=HUXme6ATYhzzt7mD&q=85&s=f0ff456ad8aa57cfedc6291d915bd330

Pipecat plugin

pipecat-slng routes your Pipecat pipeline through the SLNG gateway.

On a custom stack? Point it at the drop-in gateway and keep your code.

Try it live

Not ready to write code? Try the demos in your browser. Drop in your SLNG key and go.

Live TTS demo

Type something and hear it back. Try different models and voices.

Live STT demo

Speak into your mic and watch the transcription stream in.

Source code: github.com/slng-ai/examples

What we want you to build

Build whatever you want with voice and speech AI. A few starting points:

Voice assistants

An agent that books appointments, handles support, tutors students. Pick a task and nail it.

Accessibility tools

Speech interfaces, screen readers, audio descriptions. Make something more people can use.

Creative audio apps

Podcasts, storytelling, audio games, voice-driven art. Surprise us.

Real-time transcription

Live captioning, meeting notes, multilingual translation. Turn speech into something useful.

Prizes

Best overall project wins a Lego set. Demo what you built and impress the judges.

FAQ

How much API credit do I get?

€4 in free credits when you sign up. That covers roughly 80 hours of audio generation or transcription, so you won’t run out this weekend.

What if I run out of credits?

Come find us at the event. We’ll top you up.

Do my credits expire?

Nope. Keep building after the hackathon, polish your project, start a new one. They’re yours.

Which models should I use?

For most projects, start with:

English TTS: slng/deepgram/aura:2-en — low latency
English STT: slng/deepgram/nova:3-en — high accuracy
Spanish or Hindi TTS: slng/rime/arcana:3-es, slng/rime/arcana:3-hi
French TTS (WebSocket only): Cartesia Sonic 3 or Kugel
Multilingual STT: slng/deepgram/nova:3-multi — auto-detects 10+ languages

Full catalog on the Models page.

Can I use SLNG with my favorite framework?

Yes. It’s standard HTTP and WebSocket, so it works with anything. The Unified API follows the same URL pattern for every provider, so you can drop SLNG into an existing voice pipeline without rewriting your integration.

Do I need a paid account?

No. Sign up at app.slng.ai and you’re good to go. The free tier covers the whole hackathon.

Where can I get help?

Find us at the event, we’re around all weekend. You can also check the full docs.

​What is SLNG?

Text-to-Speech

Speech-to-Text

Voice Agents

​Fastest start: let your coding agent build it

SLNG agent skills

​Building in n8n?

​The old way, write code

​Start building

​Other languages

​Add it to your existing agent