When you share your camera in Gemini Live, you prompt with your voice and what the camera sees. Google has also added Guided Vision, which it says gives spoken descriptions of your surroundings. These prompts are our own, written for everyday use. Results vary, and Google says Guided Vision can make mistakes, so double-check anything important.
Tips for clearer answers
- Say what you are pointing at. “This is the back of a cereal box” helps more than silence.
- Ask one thing at a time. Follow up instead of stacking five questions.
- Give the goal. “I need to know if this contains nuts” beats “read this”.
- Ask for camera help. Ask to be told if the label is cut off or too dark.
- Verify the important bits. Do not rely on it for safety, medical or navigation decisions.
Prompts to try
Reading small print
“Read the ingredients on this label out loud, slowly. Tell me if any part is cut off or blurry so I can move the camera.”
Finding something nearby
“I’m looking for my keys on this table. Describe what you can see, left to right, and tell me which way to turn the camera if you can’t see the whole surface.”
Colours and patterns
“Do these two tops match? Describe the colours and patterns and give me an honest opinion in two sentences.”
Appliance dials
“This is my oven dial. Read the markings and tell me which setting is currently selected.”
New rooms and tabletops
“Give me a quick overview of this tabletop: the main objects first, then smaller details.”
Step-by-step help
“I’m assembling this shelf. Look at the parts and tell me which one to pick up next. Wait for me to say done before the next step.”
Related
- AI prompt workflow to save and improve the prompts that work
- Gemini tool page and ChatGPT vs Gemini