AI for Creative Work

🎨 The shift nobody warned you about

The conversation about AI and creativity usually goes one of two ways. Either it's all excitement — "it can paint anything!" — or all anxiety — "it's stealing from artists." Both conversations skip what's actually useful to know.

Here's the practical reality: AI hasn't replaced the need for creative vision. It's dramatically reduced the cost of the first draft. A musician who used to spend three hours making a demo can now have a rough sketch in fifteen minutes. A writer applying for a grant can get a first structure in ten. A photographer can generate a mood board in seconds instead of hours of searching.

In this module we'll look at what's actually useful across four creative domains: images, writing, music, and video. We'll try each one. And at the end there's an exercise that brings it together into something you can use for a real project you're working on right now.

🖼️ Image generation — what's actually good

Generating images has become genuinely easy. The tools below range from free browser apps to professional-grade generators. Click any card to open it — opening 3 updates your progress.

A prompt for image generation isn't a Google search. It's closer to a brief you'd give a photographer or an illustrator. The more specific you are about mood, composition, and style, the better. Try this structure:

Notice what that prompt includes: subject, setting, time of day, specific local detail, visual style, what not to include. That last part — telling it what you don't want — is often the most useful sentence in an image prompt.

✍️ Writing — beyond the office email

You already know from Module 2 that AI is good for business writing. But the same tools are surprisingly capable for creative writing — fiction, poetry, scripts, artist statements, and the kind of thing that's genuinely hard to write: grant applications.

Try this with Claude or ChatGPT right now:

The trick is the last line. You're not asking it to write for you yet — you're asking it to interview you first, so that what it writes is actually yours.

For fiction and creative writing specifically, the models that handle long, structured narrative best are Claude (for coherent multi-section pieces) and ChatGPT (for dialogue-heavy scripts and genre fiction). Both understand New Zealand context reasonably well, though you'll want to fact-check anything about te reo Māori or specific local history.

🎵 Music & audio — this one is genuinely wild

If there's one creative AI capability that reliably drops jaws in a live demo, it's music generation. You type a description. You hit generate. Thirty seconds later there's a complete song — vocals, instruments, structure — playing back. Not MIDI. Not a rough sketch. A track.

Whether you use it as a sketch tool, a production shortcut, or just for fun is entirely up to you. The tools are below.

Try this prompt in Suno right now:

🎬 Video & motion — the frontier

Video generation is moving faster than any other creative AI domain. The tools available now would have seemed impossible three years ago. They're still imperfect — hands, physics, coherent long sequences — but improving at a pace that makes specific predictions pointless.

What they're genuinely good at right now: short atmospheric clips, transitions, mood pieces, and concept visualisations. Think five-second shots, not feature films.

The honest position here: video is where you'll want to watch and wait rather than build a workflow around just yet. The tools change significantly every few months. What's useful to do now is understand what they can produce and have a rough feel for the prompting style — which is similar to image generation, just with added notes on camera movement and duration.

🔬 Google AI Studio — the creative playground

Google AI Studio deserves its own mention because it doesn't fit neatly into any one category — it's a free browser-based tool for experimenting with Gemini (Google's AI) at full power.

What makes it relevant to creative work: it's multimodal. You can drop in a photo, a sketch, a PDF, or a video clip and ask the AI to respond to it — describe it, write in its style, suggest changes, build something from it. You're not just typing prompts into a chat box; you're feeding it visual and contextual material and seeing what comes back.

• Upload a sketch of a room layout and ask for colour palette suggestions based on the architecture
• Drop in a PDF of a previous grant application and ask "what's weak about this? How would you strengthen section 3?"
• Paste a playlist and ask for lyrical themes that connect the songs — useful for writing in a particular register
• Upload a photo of a location and ask it to write an atmospheric scene set there

🎙️ NotebookLM — your documents as a podcast

NotebookLM (notebooklm.google.com) is a separate Google product that does something peculiar and genuinely useful: you upload your documents, and it becomes an expert on them. Then it can do something remarkable — generate an audio podcast discussion of what you gave it.

Two AI hosts discuss the content of your files in natural conversation — analogies, back-and-forth, key-point summaries. Upload a 40-page PDF, get a 12-minute podcast.

• Raglan use cases: turn your services brochure into a walking-tour-style audio guide; convert a business plan into a briefing for a meeting; give your portfolio to it and get a spoken artist statement draft
• What makes it different from ChatGPT: it only answers from what you gave it — not from the internet. Every answer links back to the source passage. Reliable and verifiable.
• Free, just needs a Google account. Upload PDFs, Google Docs, web pages, or YouTube links as sources.

⚖️ The honest bit — copyright and attribution

It would be dishonest to spend a whole module on creative AI without covering this. The copyright situation around AI-generated creative work is genuinely unresolved. Courts in the US, UK, and EU are actively hearing cases. New Zealand law hasn't caught up. So here's the practical guidance, without pretending there's more certainty than there is:

• Be transparent. If you use AI as a significant part of creating something — especially if you're selling it or entering it in competition — say so. The conversation is happening whether you participate or not.
• Adobe Firefly is the safest for commercial image work. It was trained on licensed Adobe Stock content. The others have less clear provenance.
• "In the style of [living artist]" is a grey area. Style itself isn't copyrightable, but using a specific artist's name in a prompt to produce commercial work will become a legal issue sooner or later.
• Most platforms grant you commercial rights to what you generate via their paid tiers. Read the terms — they change.
• Music is the most complicated. Suno and Udio are currently fighting multiple lawsuits from major labels. What you generate is probably fine for personal use. Commercial release is hazier.