Using AI for Images and Video

Getting specific pays off more here than anywhere else

Image and video generation tools respond dramatically to specificity. "A dog" produces a generic result. "A scruffy golden retriever puppy running through autumn leaves, golden-hour lighting, shallow depth of field" gives the model far more to work with.

Elements worth specifying

Subject and action: what is happening, not just what is present.
Style: photorealistic, illustration, 3D render, a particular art movement.
Lighting and mood: golden hour, dramatic shadows, soft and bright.
Composition: close-up, wide shot, from above.
Format: square, portrait or landscape, depending on where it will be used.
What to avoid: many tools support negative prompts to exclude unwanted elements.

Building a prompt step by step

This is an illustrative progression, not output from a test:

"A coffee shop." Generic and unpredictable.
"A small neighbourhood coffee shop at dawn, steam rising from a cup on a wooden counter." Now it has a subject and a mood.
"...soft morning light through the window, photorealistic, close-up shot, shallow depth of field, landscape format." Now it has style, lighting, composition and format.

Iterate on one element at a time. If the composition is right but the lighting is off, ask for a lighting change instead of regenerating from scratch. Otherwise you lose the parts that were already working. This is the same habit as in Getting Better Answers Through Iteration.

Where these tools still struggle

Hands, text inside images, and precise counts of objects (like "exactly 5 apples") remain harder for most image models. Expect to regenerate a few times, or fix those parts in an editor. Keeping the same character or product consistent across several images is also difficult unless the tool offers a reference-image feature.

Using images responsibly

Don't pass AI images off as real photos, especially of news events or real people.
Avoid generating real people's likenesses without their permission.
Add alt text describing the image for anyone using a screen reader.

Rights and disclosure: know before you use

Ownership and usage rights for AI-generated images vary by tool and by country, and the law is still developing. In some countries, including the United States, purely AI-generated material may not qualify for copyright without meaningful human authorship, so check the rules where you live. For anything commercial, read the specific tool's terms of service instead of assuming standard copyright rules apply. Some platforms and publications also require you to disclose that an image or video was AI-generated.

Related: See our dedicated guides for Photography and Video Editing.

Frequently asked

Can I use AI-generated images commercially?

Usually yes with most major tools, but terms vary — always check the specific platform's usage rights before commercial use.

Why do AI images sometimes look "off" in ways I can't quite name?

Often subtle issues with lighting consistency, anatomy, or texture that don't match real-world physics — worth a close look before using anything in a professional context.

Is AI video generation as mature as image generation?

Not yet — video generation is newer and generally requires more iteration and shorter clip lengths compared to the more established image generation space.

← Back Next: Combining Multiple AI Tools →