Node-RED text to speech with a local natural voice
Node-RED connects sensors, home automation, chat bots and business systems with a few wires on a canvas. Speech fits right in: a door sensor that announces a visitor, a flow that reads out a daily summary, a warehouse alert spoken on a shop-floor speaker.
The free SuperClonica engine runs as a small HTTP server next to Node-RED and turns text into natural speech on your own hardware. This guide shows the flow, node by node, and what to do with the audio.
What you can voice
Node-RED moves the data; SuperClonica turns the text into speech on your own network.
- Announcements from sensors and smart home events.
- Daily summaries built from other nodes.
- Alerts on factory, warehouse or office speakers.
- Messages for chat bots that send voice notes.
Choose the voice free
On the page of any voice, type a typical announcement, 30 to 100 characters, choose the language and press Voice it. In a few seconds you hear how your flow will sound, with no account.
Build the flow
If Node-RED runs in Docker too, put both containers on one network and use http://tts:8000 instead of localhost.
- Start the server on the same machine or network: docker run -d --name tts -p 8000:8000 -v superclonica-models:/models superclonica/superclonica:cpu.
- In a function or change node, set msg.payload to an object such as {"text": "Someone is at the front door", "voice": "F1", "lang": "en"}.
- Add an http request node: method POST, URL http://localhost:8000/synthesize, and return a binary buffer; an object payload is sent as JSON automatically.
- The node outputs the WAV as a buffer in msg.payload, ready for the next node.
Save or play the audio
A write file node saves the buffer as a WAV, for example /data/tts/door.wav. On a Linux machine with speakers, an exec node can play it with aplay; for smart speakers, hand the file to your home automation system or serve it over HTTP. Voice fixed phrases once and reuse the files, and voice only the parts that change. Set SUPERCLONICA_API_KEY and send it in an Authorization header if the server is reachable from other machines.
Voices and honesty
The ten stock voices of the model, M1 to M5 and F1 to F5, are free; for a voice of your own, train it in the SuperClonica studio from a recording of a speaker who agreed. Say that announcements are machine-generated where that matters, and never automate speech in a voice that imitates a real person.