Beginner10 min read

Classic Mode: Complete Tutorial

Master the fundamentals of AI image generation. Help your team guess words through perfect prompts.

Think of this as Pictionary, but instead of frantically sketching stick figures with your non-dominant hand, you're describing images to an AI. Which sounds easier until you realize the AI is very literal and your teammates only have about 10 seconds to figure out what you meant.

I watched my friend try to generate "Coffee" last week. He wrote "Brown liquid in a cup." The AI gave him muddy water in a Solo cup. Zero guesses. Brutal.

How Classic Mode Works

You get a secret word—"Coffee," "Mountain," "Robot," whatever. You describe it to an AI without actually using the word itself. The AI spits out an image based on your description. Your teammates try to guess the original word.

If someone guesses correctly, both of you get points. The faster they guess, the more points. Pretty straightforward.

The whole challenge: Make it obvious enough that people guess quickly, but interesting enough that it's not boring. This balance is harder than it sounds.

What Makes a Prompt Actually Work

After watching hundreds of rounds, I've noticed winning prompts usually have three things figured out:

Thing 1: Be Specific About Your Subject

Generic descriptions produce generic, forgettable images.

❌ "An animal"

AI could make literally anything. A fish. A weird blob creature. Who knows.

✅ "A golden retriever puppy"

Now we're getting somewhere. People can guess this.

Thing 2: Give It a Setting

Where is this thing? What's it doing? Context clues help your teammates and make the image more interesting.

Example: "playing in a sunny backyard, chasing a ball"

Adds action and environment. Much better than just "a puppy standing there."

Thing 3: Tell It What Style You Want

This is where art keywords come in. They're basically cheat codes for controlling the vibe:

"photorealistic" → makes it look like a real photograph
"oil painting" → gives it that classic painted look
"anime style" → big eyes, vibrant colors, you know the vibe
"watercolor" → soft and artistic
"3D render" → Pixar movie aesthetic

Putting It Together

Word: Dog
What worked for me:
"A golden retriever puppy playing in a sunny backyard, chasing a ball, photorealistic style, joyful expression"

Specific breed + action + setting + style = teammates guessed it in 4 seconds.

The Speed vs Creativity Dilemma

Here's the tension: you want people to guess fast (more points), but you also don't want to make a boring image that gets zero reaction from the group.

I've learned it depends on the word.

For Simple, Concrete Words: Just Be Obvious

If your word is something like "Tree" or "Pizza," don't overthink it. Make it unmistakably that thing:

Word: "Tree"
→ "A tall oak tree in a forest, green leaves, photorealistic"

No tricks, no metaphors. Just a tree.

Word: "Pizza"
→ "A pepperoni pizza on a wooden table, melted cheese, top view"

Everyone knows what this is instantly. Easy points.

For Abstract Words: Get Creative

When your word is something like "Freedom" or "Mystery," you can't just show the literal word (that's cheating). Time to get visual and poetic:

Word: "Freedom"
→ "A bird breaking free from a cage, wings spread wide, sunrise background, dramatic lighting"

Symbolic representation. People get it.

Word: "Mystery"
→ "A foggy detective's office, silhouette in shadows, film noir style"

Evokes the concept through mood and setting.

Mistakes I've Made (So You Don't Have To)

Some lessons learned the hard way:

Using the actual word in your prompt

Word: "Sunset" → Prompt: "A beautiful sunset over the ocean"

Yeah, this gets you disqualified or zero points. Happened to me once. Embarrassing.

Better: "The sun setting below the horizon, orange and pink sky, silhouettes of trees"

Being way too vague

Prompt: "Something cool"

The AI literally doesn't know what to do with this. Neither do your teammates when they see the result.

Writing a novel instead of a prompt

Prompt: "A fluffy white cat with blue eyes wearing a tiny red bow tie sitting elegantly on a velvet cushion in a Victorian-era parlor with ornate furniture and paintings on the walls while sunlight streams through lace curtains..."

The AI stopped reading halfway through. So did I, honestly.

Better: "A fluffy white cat with a bow tie sitting on a velvet cushion, Victorian parlor, elegant"

Using "not" instructions

Prompt: "A dog, not scary, not dark, not aggressive"

AI sees the words "scary," "dark," "aggressive" and goes "got it, making nightmare dog."

Better: "A friendly happy dog, bright sunny lighting, playful"

Stuff That Works Better Than You'd Think

Camera Angle Keywords

Want the AI to show your subject from a specific angle? These work surprisingly well:

"close-up shot" → zooms in on the main thing
"wide angle view" → shows the whole scene
"from above" / "top-down view" → bird's eye perspective
"eye level" → like you're standing there looking at it

Lighting Makes Everything Better

Adding lighting keywords can completely change how an image feels:

"soft natural lighting" → looks pleasant and gentle
"dramatic lighting" → adds intensity
"golden hour" → that warm sunset glow everyone loves
"studio lighting" → professional and clean

Different AI Models Have Different Vibes

Your lobby might be using different AI models. Here's what I've noticed:

DALL-E 3: Likes conversational prompts. Write like you're talking to someone.

Example: "A cozy coffee shop on a rainy day with people reading books"

Flux Dev: Better with artistic stuff. Works well with art style keywords.

Example: "Mountain landscape, oil painting style, dramatic clouds"

Flux Schnell: Fast but simple. Keep it straightforward.

Example: "A red car on a highway, photorealistic"

Try It Yourself

Here are some words to practice with. Take 20 seconds per word and write a prompt:

Word: "Ocean"
Challenge: Everyone's first thought is "blue water and waves." How do you make yours stand out?
Word: "Robot"
Challenge: Make it clearly a robot, but not just a generic metallic humanoid.
Word: "Summer"
Challenge: It's abstract. What visual says "summer" without being cliché?

👉 Here's what I'd write (click to see)

Ocean: "Underwater view looking up at the surface, sunlight beams penetrating blue water, school of fish swimming, photorealistic"

Different perspective than the usual surface view. Makes it memorable.

Robot: "A friendly robot assistant in a modern kitchen, helping with cooking, smooth white design, blue LED eyes, 3D render"

Gives it personality and context. Not just standing there.

Summer: "Children running through a sprinkler in a backyard, water droplets sparkling in sunlight, laughter and joy, warm tones"

Captures the feeling through a specific memory most people have.

Quick Cheat Sheet

When in doubt, use this structure:

[What it is] [doing what] in [where], [art style], [lighting]

Example: "A red sports car driving on a coastal highway, ocean view in background, cinematic photography, sunset lighting"

This formula works for probably 70% of words. Master it first, then get fancy.

Alright, enough theory. Time to actually play. Create a Classic mode lobby and see if your prompts hold up against real humans.

Spoiler: your first few rounds will probably be terrible. That's normal. We've all been there.