The new creative “starter kit” (and why it’s not cheating)
A lot of folks show up to NH AI Meetup with the same nervous question: “Is this… allowed? Like, am I faking it if I use AI to make art?”
My take: using AI isn’t cheating, it’s just a different tool. Cameras didn’t kill painting, and spreadsheets didn’t kill math. AI is closer to a sketchbook that talks back. You’re still making choices: what you ask for, what you reject, what you keep, and how you stitch it into something that feels like you.
The best part is you don’t need to be an expert. You just need a workflow that keeps you in control. Let’s walk through three lanes—art, music, and stories—with concrete prompts and small “recipes” you can try tonight.
A simple mindset: AI is your intern, not your boss
Before the tools, one mindset shift: treat the model like an eager intern. It’ll produce a lot, fast. Some of it is brilliant. Some of it is nonsense with confidence. Your job is to steer.
A good loop looks like:
- Intent: What are you trying to make, and for who?
- Constraints: Style, length, mood, instruments, color palette, audience.
- Generate: Get 5–20 rough options, not “the perfect one.”
- Curate: Pick the best 10%.
- Edit: Fix the weak parts yourself (or with another AI pass).
- Finish: Export, format, polish, share.
That’s it. The magic is mostly in steps 2 and 4, honestly.
Visual art: go from vague idea to a usable image
Tool options (non-exhaustive, because this space changes weekly)
- Midjourney (Discord-based, strong aesthetics)
- Stable Diffusion (local or hosted, very customizable; look for SDXL)
- DALL·E / image tools in ChatGPT (easy iteration, good for quick concepts)
- Canva / Adobe Firefly (friendly UI, practical for marketing-ish work)
If you’re brand new, start with whatever feels easiest to access. If you’re the tinkering type, Stable Diffusion can become a full hobby by itself.
Prompting that doesn’t feel like spellcasting
People think prompts have to be these arcane strings of adjectives. Nah. What works is being specific about the scene and then giving a couple style hints.
Try this structure:
- Subject: who/what
- Scene: where, what’s happening
- Mood/lighting: soft morning light, neon night, foggy
- Medium: watercolor, 35mm photo, ink sketch, 3D render
- Composition: close-up, wide shot, overhead
- Constraints: “no text,” “no watermark,” “minimal background”
Example prompt:
“A cozy corner desk in a New England cabin, a laptop open with code on screen, a mug of tea, pine trees visible through a frosty window, soft morning light, warm color palette, photorealistic, 35mm lens look, shallow depth of field, no text, no logos.”
Now the practical trick: ask for variations in batches. Don’t marry the first image. Generate 8–16, pick 2, then iterate.
A quick mini-tutorial: consistent characters (without losing your mind)
If you want the same character across images (for a comic, a brand mascot, whatever), you can:
- Describe them with a “character card” (hair, clothing, age, vibe) and reuse it.
- Use reference images if your tool supports it.
- In Stable Diffusion land, people use LoRA or IP-Adapter for consistency, but that’s optional.
Character card example:
“Character: Maya, late 20s, short curly black hair, round glasses, green winter jacket, friendly but focused expression.”
Then reuse that in every prompt. It won’t be perfect, but it’ll get you 80% there, and your edits can do the rest.
Music: generate a track, then make it yours
Music AI can feel like wizardry the first time you hear it. The important thing is to treat it like a draft, not your final song.
Tools you’ll hear people mention
- Suno (generate songs with vocals; very approachable)
- Udio (also song generation; strong styles)
- AIVA / Soundraw (more “background music” oriented)
- Ableton/Logic + AI helpers (for folks who already have a DAW)
Prompting music: think like a director
Instead of listing 40 genres, pick a few anchors:
- Genre + era: “indie folk, 2010s”
- Tempo: “95 BPM, laid-back”
- Instrumentation: “acoustic guitar, brushed drums, upright bass”
- Mood: “hopeful, crisp winter morning”
- Structure: “intro-verse-chorus-verse-chorus-bridge-chorus”
Example:
“Indie folk track, 95 BPM, acoustic guitar + upright bass + brushed drums, warm and hopeful, sounds like driving through New Hampshire after a snowstorm, intro-verse-chorus-verse-chorus-bridge-chorus, no vocals.”
Then generate a few. Pick the one with the best core idea (melody, groove, chord movement).
The part beginners skip: editing
Even if you don’t “know music,” you can still do meaningful edits:
- Trim the intro so it hits faster.
- Fade in/out cleanly.
- Layer simple sounds on top (a shaker loop, a piano chord pad).
- EQ: cut some muddy low-mids if it sounds like it’s under a blanket.
Free tools like Audacity can handle basic trims and fades. If you want one step up without going full producer, try Reaper (cheap, generous trial) and watch a 10-minute tutorial. That’s enough.
Also: if you’re using vocal generation, double-check the tool’s rules about commercial use and voice styles. Some platforms will let you make “in the style of…” stuff, but you really don’t want to accidentally wander into imitation that feels too close.
Stories: co-write instead of “generate me a novel”
Text models are fantastic writing partners, but they’re also… kind of lazy. If you ask for a whole story, you’ll often get something that reads like a Wikipedia plot summary wearing a trench coat.
The fix is to break the job into pieces.
A practical story workflow (I use this a lot)
- Premise in 1 sentence
- Character wants + fear
- 3-act outline (bullets)
- Write one scene at a time
- Revision pass focused on voice
Prompts that work:
Premise + tone
“Give me 10 short story premises (1–2 sentences each). Set in a small New England town. Tone: cozy but eerie. No zombies, no serial killers. Make them original.”
Character anchors
“Pick one premise and create two main characters. For each: what they want, what they’re afraid of, and one weird habit. Keep it grounded, not cartoonish.”
Scene drafting
“Write Scene 1 in close third-person from Maya’s POV. Show, don’t explain. Keep it under 900 words. End the scene with a small unsettling detail, nothing dramatic.”
Then you, the human, do the important part: you rewrite sentences to sound like you, you cut the fluff, you add a real sensory detail you remember from your own life. (There’s always one: the smell of wet wool gloves, the way snow squeaks under boots at night.)
A tiny trick for better dialogue
AI dialogue often feels a bit too polished. You can push it toward real speech:
“Rewrite this dialogue to sound like two people talking in real life: shorter sentences, occasional fragments, mild interruptions, not everyone says the perfect thing.”
It’s a weird prompt, but it works.
Putting it together: one weekend project idea
If you want a low-pressure challenge, do this:
- Friday night: generate 12 images for a children’s-book style character.
- Saturday: pick 6 images that feel coherent and build a tiny story around them (6 scenes, 200–400 words each).
- Sunday: generate a simple instrumental theme song, trim it, and attach it to a slideshow video.
Congrats, you made a mini picture book + soundtrack. That’s a real creative artifact, not a tech demo.
A few real-world cautions (the unsexy but important stuff)
- Copyright and licensing: different tools have different terms. If you want to sell your work, read the usage policy. It’s boring. Still do it.
- Training data and ethics: folks feel differently about models trained on public art. If you’re sharing publicly, be transparent about your process.
- Personal data: don’t paste private info into random web tools. Treat prompts like they might be stored.
- Taste matters: AI can generate endless content, but it can’t decide what’s worth making. That’s on you.
Bring your experiments to the meetup
One of the best things about a community like NH AI Meetup is seeing what other people are building. Bring a half-finished track. Bring three versions of the same image prompt and ask which one reads better. Bring the weird story opening you can’t quite land.
You don’t need to be an expert to make art, music, or stories with AI. You just need a point of view, a little patience, and a willingness to iterate. The tools are loud, but your taste is the quiet superpower.