Make images and talk to AI by voice

What you will learn

• How to describe an image you can use: subject, setting, style, framing and size
• How to improve it one change at a time, and why you add the words yourself
• How to check whether you may use an image
• How to practise a conversation by voice, and check it afterwards

The idea

Many AI tools can make pictures and hold spoken conversations. For images, examples are ChatGPT, Gemini and Meta AI in a chat, or design tools such as Canva AI and Adobe Firefly. For voice, ChatGPT, Gemini, Microsoft Copilot and Meta AI are examples. Each tool page shows what we have confirmed about it, with the official source.

A short request such as "a picture for my slide" gets an average picture. Describe five things: the subject, the setting, the style (photo, flat illustration, line drawing), the framing (close-up or wide, and where to leave space for a title) and the size (wide for a slide, tall for a poster).

Treat the first image as a draft and change one thing per round, such as "same image, plain white background", so you see what each change did. Words inside a generated image can still come out misspelled, unclear or in the wrong place, even in newer tools. Ask for no text and add titles and labels yourself.

Before you use an image, check the tool's terms and your company's or school's rules. Leave out real people, logos and brands unless you have permission, and never pass a generated image off as a real photo.

Voice helps you practise a conversation, such as an interview or a difficult call, think out loud, or work hands-free. Say your context first: who it should play and what you want to practise. Often you can interrupt simply by talking. Afterwards, read the transcript where your tool keeps one, and check any facts. Voice chats are data too and may be saved like typed chats, so check your privacy settings.

Example

Maria, an office administrator, needs a title slide picture for a talk on new recycling rules. "Make a picture about recycling for a slide" gives her a busy globe with the word "RECYCLNG" across it. Her second try:

Create an image for a presentation slide.
Subject: three recycling bins, blue, green and grey.
Setting: a bright office kitchen.
Style: flat illustration, soft colours.
Framing: bins on the right, empty space on the left for a title.
Size: wide, 16:9.
No words, letters or logos in the image.

The bins come out small, so she changes one thing: "Same image, larger bins, keep everything else." She types the title into the slide herself and checks her company's guidance on AI images.

Before a difficult call with a supplier, she opens voice mode and says:

You are a supplier whose delivery is two weeks late. I am the customer.
Push back a little, keep answers short, and after five minutes tell me
which of my points were unclear.

She interrupts when answers run long, names no real company, and reads the transcript afterwards.

Practice now

1. Pick one image you need this week: a slide illustration, a poster background or a simple picture that explains a process.
2. Fill in the image part of the template below and generate it.
3. Change one thing per round, for up to three rounds, then add any words yourself.
4. Check the tool's terms and your company's or school's rules.
5. Practise one conversation by voice for five minutes. Say your context first.
6. Read the transcript if your tool keeps one, and note one thing to keep and one to check.

At the end you should have one image you can use and notes from one voice conversation.

Copy this template

Image:
Create an image for [a slide, a poster or another use].
Subject: [what is in the image].
Setting: [where it is].
Style: [for example photo, flat illustration, line drawing].
Framing: [close-up or wide, and where to leave empty space].
Size: [wide, tall or square, or a ratio such as 16:9].
No words, letters or logos in the image.

Voice practice (read it aloud, or type it and switch to voice if your tool allows):
I want to practise [conversation]. You are [role]. I am [my role].
Keep your answers short. At the end, tell me what went well, what was
unclear, and anything you are unsure about.

Asking AI to flag uncertainty does not make its answer correct. Keep checking the parts that matter.

Check the output

• Does the image show your subject in the style and shape you asked for, with room for text?
• Is it free of garbled words, logos and recognisable real people?
• Do the tool's terms and your company's or school's rules allow this use?
• Does the transcript match what you heard, and are its facts right?
• Did you keep names and private details out of both?

Common mistakes

• Changing everything at once. You cannot tell what helped. Change one thing and keep the rest.
• Leaving words in the image. A misspelling is easy to miss and hard to fix. Add text yourself.
• Assuming you may use any image you made. Rules differ between tools, companies and schools. Check first.
• Starting a voice chat with no context. You get small talk. Say who it plays and what you are practising.
• Trusting what you heard. Spoken answers sound sure and are gone in seconds. Read the transcript where there is one.