How to Use AI on Your Phone: Voice, Camera and Gemini Live

By Ravindra Bagale, trainer · LinkedIn · Last updated: 19 October 2026

Generative AI course · Chapter 5 · Estimated reading time: 14 minutes

1. Suppose we are at a hardware shop with a long bill

Suppose we are Amit. We are getting a small bathroom repaired at home in Kolhapur. The plumber bought materials from a hardware shop, and gave us a printed bill with 8 items. We want to check the total and keep the items in a sheet.

Typing 8 items with prices on the phone will take 10 minutes. Instead:

1. We open the ChatGPT or Gemini app.
2. We tap + (or the camera icon) and take a photo of the bill.

3. We type one line:

   Turn this bill into a table with columns: Item, Quantity, Rate, Amount. Then add the total.
   

4. In a few seconds we have a table. We check it against the paper bill, and we are done.

This is the power of AI on a phone: show it, say it, and then check it. In this chapter we do this properly, step by step.

2. Three ways to use AI on a phone

All 3 ways work in the ChatGPT and Gemini apps. Claude and Copilot apps also have voice and photo upload. Some live features depend on your plan and country, as you will see below.

Install the right app. In the Play Store or App Store, check the publisher: OpenAI for ChatGPT, Google LLC for Gemini, Anthropic for Claude, Microsoft Corporation for Copilot. Fake look-alike apps exist.

3. Worked example 1: a photo of a bill becomes a table

What you need: any printed shop bill (grocery, hardware, medical store). Cover or fold away any phone number, GST number of a person, or name you do not want to share.

1. Place the bill flat on a table, in good light, with no shadow of your phone on it.
2. Open the ChatGPT app (or Gemini).
3. Tap + next to the message box, then choose Camera (or Photos if you took the picture already).
4. Take a straight photo that shows all the lines of the bill.

5. Type this prompt and send it with the photo:

   Role: You are a careful accounts assistant.
   Task: Read this shop bill and copy it into a table.
   Format:
   - Columns: Item, Quantity, Rate (₹), Amount (₹).
   - Copy the numbers exactly as printed. If a number is not clear, write "unclear".
   - After the table, write the total of the Amount column, and the total printed on the bill.
   - If the two totals are different, say so.
   

Expected output (from a made-up hardware bill; yours will show your bill):

Total of the Amount column: ₹1,956
Total printed on the bill: ₹1,956
The two totals match.

6. Check it yourself. Hold the bill next to the phone and compare every line. Then copy the table into Google Sheets or Excel and type =SUM(D2:D9) in the cell under the Amount column. For this made-up bill, the result is 1956.

7. Follow-up idea: Which item is the most expensive, and what part of the total is it? (For this bill: the wall tap, ₹650, about 33 % of ₹1,956.)

A note on free limits: photo uploads count toward a daily limit on free plans. For example, OpenAI's help page says free ChatGPT users can upload 3 files a day (checked October 2026). If you hit the limit, use Gemini for that photo, or try again later.

4. Voice: speak your prompt

There are two kinds of voice:

1. Dictation: you speak, and the app types your words into the message box. You check the text and then send it. Look for the microphone icon.
2. Voice chat: a back-and-forth spoken conversation. The AI listens and talks back. Look for the Voice icon (often a sound-wave shape).

ChatGPT Voice (checked on OpenAI's help page, October 2026)

1. Open the ChatGPT app.
2. Tap the Voice icon in the message bar.
3. If asked, allow the app to use your microphone.
4. The first time, choose a voice you like.
5. Start speaking. You can interrupt it, like a normal phone call.
6. Tap the exit control to end. A text copy of the conversation is added to your chat.

Good to know:

• Free users get limited voice use. Paid plans get more.
• Under Settings → Voice you may see Live, Advanced and Standard. OpenAI says Live does not support video or screen sharing. Video and screen sharing are available to eligible paid subscribers in the phone app, using Advanced.
• You can change the voice language under Settings → Voice → Language.
• OpenAI says audio clips from voice conversations are kept with the chat for 30 days.

5. Worked example 2: practise English speaking for an interview

Suppose we are Pooja, a BSc graduate from Satara. We have a phone interview for a customer support job at an e-commerce company next week. We are nervous about speaking English fluently. Voice chat is a safe place to practise.

1. Open ChatGPT (or Gemini, Claude or Copilot) and start voice chat.

2. Say this, slowly and clearly:

   Please act as an interviewer for a customer support job at an online shopping company.
   Ask me one question at a time. Wait for my full answer.
   After each answer, tell me two things I did well and one thing to improve, in simple English.
   Start with: Tell me about yourself.
   

3. Answer each question in your own words.

4. After 4 or 5 questions, say: Stop the interview. Give me a short summary of my 3 biggest improvement points.
5. End the voice chat and read the text copy in the chat. Copy the improvement points into your notes.

What it usually sounds like (yours will be similar, not the same):

Interviewer: Tell me about yourself.
You: (your answer)
Interviewer: Thank you. Two good things: you spoke clearly and you mentioned your
computer skills. One thing to improve: you said "I am having 1 year experience";
a better way is "I have one year of experience". Next question: how would you
handle a customer whose order arrived damaged?

Practise this 10 minutes a day for a week. Speak, listen to the feedback, and try the same answer again.

6. Gemini Live: talk while showing your camera or screen

Gemini Live is a live voice conversation in the Gemini phone app. While you talk, you can turn on your camera or share your screen, and Gemini talks about what it sees. Google made camera and screen sharing in Gemini Live free on Android and iPhone in May 2025.

What you need (from Google's help page, checked October 2026):

• The Gemini app on an Android phone or an iPhone. Gemini Live is not in the Gemini website.
• To be signed in to the Gemini app with your Google account.
• To be over 13 (or the minimum age in your country).
• For screen sharing, notifications for Gemini Live must be turned on.
• On Android, Google lists camera and screen sharing for phones with Android 10 or newer and at least 2 GB of RAM.

Start Gemini Live with the camera:

1. Open the Gemini app.
2. Tap Live (at the bottom of the screen).
3. Tap Camera.
4. Point the camera at the thing you want to ask about.
5. Start talking.
6. To switch between the front and back camera, tap Switch to rear or front camera.
7. Tap Camera again to stop, and end Live when you are done.

Share your screen:

1. Open the Gemini app and tap Live.
2. Tap Turn on screen sharing and follow the steps on the screen.
3. Go to the app or page you want help with, and ask your question aloud.
4. To stop, go back to Gemini and tap Turn off screen sharing, or swipe down from the top of the screen and tap Stop sharing.

On Android, you can also long-press the power button (if Gemini is your assistant) and tap Share screen with Live.

The camera turns off by itself if you leave the Gemini app, lock the screen, or put Live on hold.

7. Worked example 3: ask about a plant from a photo

Suppose a new plant has come up in a pot on our balcony in Nashik. We want to know what it is and how to look after it.

Option A: a photo in a normal chat (works in ChatGPT, Gemini, Claude and Copilot)

1. Take a clear photo of one full leaf and the stem, in daylight.
2. Take a second photo of the whole plant.

3. Send both photos with this prompt:

   Task: Tell me what plant this probably is.
   Format:
   1. Your top 2 guesses, with how sure you are (high, medium or low).
   2. What in the photo made you guess this (leaf shape, colour, stem).
   3. Basic care: sunlight and water, in 3 short points.
   4. What I should check to be sure.
   Do not tell me whether it is safe to eat or use as medicine.
   

Expected output (shortened, for a made-up example):

1. Most likely a money plant (pothos), high confidence. Second guess: a philodendron, low.
2. Heart-shaped glossy leaves, green with yellow streaks, a climbing stem with small roots.
3. Care: bright light but not direct afternoon sun; water when the top 2 cm of soil is dry;
   do not let water stand in the pot.
4. To be sure, compare with photos on a trusted plant website or ask a nursery.

Option B: Gemini Live with the camera

1. Open Gemini, tap Live, then Camera.
2. Point at the plant and say: What plant is this? Look at the leaves closely.
3. Move the camera slowly to show the leaf, the stem and the soil.
4. Ask a follow-up aloud: How often should I water it in October in Maharashtra?

8. Other phone tools you may already have

9. Safety rules for voice, camera and screen sharing

Before you turn on a camera or share a screen, look at what is visible. When you share your screen, your full screen is shared, including pop-up notifications.

Never show or say:

1. OTPs, UPI PINs, passwords or bank app screens.
2. Aadhaar, PAN, passport or any ID card.
3. A screen with other people's chats or photos.
4. A child's face, school name or uniform to an AI tool.
5. Medical reports of yourself or family, unless it is a tool your doctor or hospital gave you.

Before you start:

1. Close banking, UPI and chat apps.
2. Turn on Do not disturb so notifications do not pop up while sharing.
3. Point the camera only at the object you are asking about.
4. When you finish, check that the camera and screen sharing are off.

10. Beginner mistakes and quick fixes

11. Try at home

Recap

Pick up one bill from your table right now and turn it into a sheet. Five minutes, and you will never type a bill again.

Previous chapter: How to Use ChatGPT Step by Step: Chats, Follow-ups, Memory and Projects

Next chapter: Few-Shot Prompting: Teach AI with Examples (Copy-Paste Prompts)

Ravindra Bagale, trainer: linkedin.com/in/ravindra-bagale. The people, the hardware bill, the interview and the plant on this page are made up for learning. App names, buttons and limits were checked on the official help pages on 19 October 2026 and can change. AI answers change every time, so your output will be similar, not the same.