OpenClaw Camera Application Tutorial

Before starting, please complete [01-Peripheral Hardware Configuration], [02-OpenClaw API-KEY Configuration], [03-OpenClaw Model Switching], [04-OpenClaw Prompts], [05-AI Voice Interaction Configuration], and [06-Three Conversation Scheme Configuration and Testing] in the Preparation for Use directory. This document assumes the basic environment is already prepared and only expands on camera-related content.

1. Tutorial Objectives

This tutorial demonstrates OpenClaw's natural language control capabilities through CSI camera peripherals. After completing this tutorial, readers will be able to:

• Understand the basic idea of using a camera for photography and visual analysis in OpenClaw

• Interact with OpenClaw via three methods: WhatsApp, TUI, and voice code

• Complete four camera cases: photo archiving, scene description, OCR recognition, target localization and face detection

• Understand the complete pipeline from "user proposes visual task" to "image capture" and "result return"

2. Hardware Preparation

• OpenClaw main board

• CSI camera module

• Camera ribbon cable

• [Optional] Microphone module

• Wiring diagram

3. Software Preparation

For international version:

If you are using WHATSAPP and want to upload images directly to WhatsApp, you need to fill in the conversation ID of your newly created WhatsApp conversation in the "allowFrom" path below.

• Open your newly created WhatsApp conversation to see the conversation ID.

• Fill this into "allowFrom":

• Enter the following command in the terminal to take effect:

openclaw gateway restart

4. Testing Camera Application

Before starting this tutorial, please complete a basic openclaw tui conversation self-check first; if you haven't done so, see [05-Three Conversation Scheme Configuration and Testing]. Also, please confirm that the visual model has been configured, refer to [02-OpenClaw API-KEY Configuration] and [03-OpenClaw Model Switching]. After passing, proceed to the camera-specific test.

Enter the following command in the terminal to enter TUI:

xxxxxxxxxx
openclaw tui

You can try entering the following content in the terminal:

• Help me take a photo

• Recognize the text in this image

• Find the cup in the frame (since the model will contact context, remind it to retake the photo, otherwise it might use the first photo)

• Retake a photo and organize the text inside into a table

5. Three Schemes for Conversing with OpenClaw

If you have selected the AI voice module + speaker, Feishu dialogue, tui, and WhatsApp dialogue need to broadcast the content of openclaw's reply. You can run the following program on the terminal. The AI voice module dialogue does not need to be run repeatedly

• Modify whether to enable broadcast parameter

xxxxxxxxxx
/home/sunrise/voice_to_openclaw/.env

xxxxxxxxxx
ENABLE_TTS=true  # This parameter controls whether to broadcast. After modifying and saving, run the following program in the terminal.

• Enter in terminal:

xxxxxxxxxx
cd ~/voice_to_openclaw
source .venv/bin/activate
python -m src.openclaw_session_tts

5.1 WhatsApp Scheme

For WhatsApp access steps, please refer directly to [06-Three Conversation Scheme Configuration and Testing]. This section only retains conversation examples suitable for the camera tutorial.

You can restart direct conversation, for example:

• See what is in front

• Recognize the text in the image

5.2 TUI Scheme

If you have completed the TUI self-check in Part 6, you can continue the conversation by entering the following command in the terminal:

xxxxxxxxxx
openclaw tui

Then you can directly input camera-related instructions, for example:

• Take a photo of the current scene

• Describe the scene in the lens

• Detect if there are any faces

You can add TUI conversation screenshots here.

5.3 Voice Code Scheme

xxxxxxxxxx
Requires selection of AI voice interaction module + speaker

For voice module installation, environment configuration, and basic wake-up process, please refer directly to [04-AI Voice Interaction Configuration]. After wake-up, you can directly speak control content, for example:

• Help me take a photo

• See what is in front

• Help me read the text in the frame

  

You can add voice wake-up and voice control screenshots here.

6. Comprehensive Cases

Note: The following are conversation demonstrations conducted via openclaw tui. You can also choose WhatsApp or voice methods to complete them.

6.1 Case 1: Photo Archiving

Experiment Objective

Let OpenClaw complete the most basic photography function first and save the image.

Effect Description

After execution, a photo will be generated locally, usually keeping both the current photo file and the latest.jpg alias for the most recent photo. This case is suitable for verifying the image capture pipeline.

Example Instructions

• Help me take a photo

• Capture the current scene

• Take a photo now for record

6.2 Case 2: Scene Description

Experiment Objective

Let OpenClaw provide a brief description of the current scene after taking a photo.

Effect Description

OpenClaw can first complete the photography, then call the visual analysis script to make a one-sentence description of the image content, such as identifying desks, people, food, or common objects.

Example Instructions

• See what is in front

• Help me describe the current scene

• Take a photo and tell me what is mainly in the lens

6.3 Case 3: OCR Table Recognition

Experiment Objective

Let OpenClaw recognize text in the image and organize it into a table.

Effect Description

This case is suitable for photographing labels, cards, packaging, or simple documents, and then extracting the text. It can output plain text or continue to organize it into a table.

Example Instructions

• Recognize the text in this image

• Take a photo and organize the text into a table

6.4 Case 4: Target Localization and Face Detection

Experiment Objective

Let OpenClaw locate specific targets in the image, or detect if there are any faces in the frame. If a face is detected, the selected photo will default to:

xxxxxxxxxx
/home/sunrise/.openclaw/media/csi-camera/annotated/latest-face.png

Effect Description

This type of task is not just about "seeing," but further finding the position of target objects in the image, or recognizing faces and outputting annotation results. It is suitable for interactive displays and simple visual tasks.

Example Instructions

• Find the cup in the frame

• See if there are any faces

• Help me locate the mobile phone on the right side of the desk