OpenAI open-sources Plant Talk: Connect ChatGPT to make your potted plants talk

OpenAI recently open-sourced a project called Plant Talk, allowing houseplants to "speak" through ChatGPT. The minimum setup only requires a computer with a camera and microphone, while the advanced version can connect to an Arduino soil moisture sensor and a light sensor for more accurate plant health data.

(Previous context: OpenAI announced the "Patch the Planet" project, debugging 19 well-known open-source projects including cURL, Python, PyPI, etc.) (Background: OpenAI's first AI hardware product leaked, smart speaker can recognize faces, observe, and help you shop)

OpenAI officially open-sourced the Plant Talk project on GitHub this week under the Apache 2.0 license, available for anyone to use. The official description is "a potted plant you can talk to using ChatGPT." Users can ask the plant, "How have you been lately?" or "Do you need water?" and the plant will respond with voice.

Minimum Setup vs. Advanced Setup: The Gap Is Larger Than Imagined

The hardware requirements for Plant Talk are cleverly designed, but there's an important gap.

The minimum setup—what the official description calls the "anyone can do it" version—only requires a computer with a webcam and microphone, plus an OpenAI API key. The camera captures the plant's condition, and the system periodically performs structured "observations" to assess the plant's health, then uses the Realtime API to make the plant speak.

The advanced setup is for advanced builders: adding an Arduino-compatible microcontroller, a capacitive soil moisture sensor, an LM393 light sensor, jumper wires, and a breadboard allows reading actual environmental data like humidity, light, and CO₂ levels. The Arduino collects plant data, sends it to ChatGPT to generate care suggestions, and then reads them out via text-to-speech.

The key is this sentence in the official README: Without Arduino, the system only relies on the camera for judgment, and accuracy is "much less accurate."

Simply put, guessing whether a plant needs water with a camera and directly measuring values with a soil sensor are two entirely different levels of precision. The former is more like a demo for show, while the latter is a truly practical configuration.

Codex Desktop Is the Real Star of This Demonstration

The user interface of Plant Talk is designed as a "small control room for the plant": a dashboard displays plant status, Start camera activates the camera, Observe now triggers an observation, Connect Arduino connects sensors, Talk to plant starts voice dialogue, and Open ambient mode enters a full-screen conversation view. Most users are advised to enable Ambient Mode, creating a context between the screen and the actual plant.

Users only need to open the Codex Desktop App, type a single line—"Help me make Plant Talk https://github.com/openai/planttalk"—then let Codex read the entire repo, guide you through each setup step, and build the entire project.

The plant is the facade, but Codex is the product. This is what OpenAI wants to demonstrate with this open-source project.