Fablefjorrhythmfx
Chapter 2 · Create

Chapter 2: Images, Audio and Voice Tools

AI can now draw, clean up photos, write out what was said in a meeting and even imitate a voice. The creative jobs are fun. The consent questions are serious.

3tool families
4consent steps
9 minreading time

2.1 Image generators and photo editors

Image generators such as Adobe Firefly, OpenAI's DALL·E (inside ChatGPT), Midjourney and Canva's Magic Studio turn a written description into a picture. They are excellent for mood boards, birthday card illustrations, placeholder art for a presentation or visualising a room layout before you paint. Phone galleries now include editing features too, such as Google Photos' Magic Eraser, which removes a stray bin or passer-by from the background of a holiday shot.

Where they slip: text inside images is often garbled, hands and reflections can look wrong, and anything requiring real-world accuracy, such as a map, a diagram of wiring or a product photo for sale, should not be generated. A buyer has a right to see the actual item.

2.2 Fair to make, fair to share

Generators can imitate styles and faces. A sensible rule: do not create images of real, identifiable people without their permission, and do not pass off generated work as a photograph. Several platforms and camera makers support Content Credentials, a standard from the Coalition for Content Provenance and Authenticity (C2PA) that attaches a record of how an image was made. When you share AI images publicly, a simple "AI-generated" caption is honest and increasingly expected. Copyright rules for AI output differ between countries and are still being settled, so check the tool's terms before using images commercially.

2.3 Transcription and captions

Speech-to-text has quietly become one of the most dependable AI features. Otter.ai, apps built on OpenAI's Whisper model, Google's Live Transcribe on Android and Live Captions on iPhone and Mac can turn speech into text in close to real time. These are a gift for people who are deaf or hard of hearing, for students reviewing lectures, and for anyone who thinks better out loud.

Check proper names, numbers and technical terms, which are the usual casualties. And ask before you record: in many places recording a conversation without the other people knowing is unlawful, and it is impolite everywhere.

  1. 1
    Announce it. Tell everyone present that a transcript is being made and why.
  2. 2
    Offer an opt-out. If anyone objects, take notes by hand instead.
  3. 3
    Store it carefully. Keep transcripts in the same place you would keep the meeting's documents, not a personal app.
  4. 4
    Delete on schedule. Remove recordings once the notes are checked and shared.

2.4 Voice cloning and describing the world

Services such as ElevenLabs can produce a synthetic voice from a short sample. Legitimate uses include narrating your own videos or helping people who are losing their voice through illness to bank it. The same technology powers scams: a call that sounds like a grandchild in trouble, urgently asking for money. Consumer protection bodies such as the US Federal Trade Commission have warned about exactly this.

Safety noteAgree a family code word that a real relative would know and a caller would not. If an urgent money request arrives by phone, hang up and call the person back on a number you already have. Never clone someone else's voice without clear, written consent.

On the helpful side, apps such as Google Lens and Be My Eyes can describe a photo, read a label aloud or translate a sign, which is a real independence boost for blind and partially sighted people. Treat descriptions of medication labels or allergens as a prompt to double-check, not a final answer.

Portrait of Daniel Reyes, editor of Chapter 2
Daniel Reyes
Chapter editor · reviewed February 2026