QWE AI Academy Leonardo AI: you can generate your first image in under 60 seconds, or spend 20 minutes adjusting settings that drain your free tokens before you realize what happened.
I burned through 80 tokens on a single “high-quality” generation before understanding what Alchemy mode actually costs. Tutorials skip this part. They show you the pretty interface, the model dropdown, the Generate button. They skip the moment you check your token balance and see 12 left out of your daily 150.
The Real Choice: Web Interface or Wait for Magic
Leonardo AI runs in your browser – no Discord, no command-line confusion. Clean dashboard. Actual buttons.
Method A: Standard Web Workflow
Sign up. Pick a model. Type a prompt. Hit Generate. Download your image.
90% of users do this. Fast. Works.
Method B: API-First Approach
Developers skip the web app, go straight to Leonardo’s API. Different pricing (API credits, not web tokens). You can automate hundreds of images. But unless you’re building an app, this is overkill.
Method A wins. The API launched in 2026 with an opaque credit-based pricing model. For learning? Web interface every time.
Why Leonardo Exists (And Why It’s Not Midjourney)
Leonardo AI: Australian platform, Canva acquired it in 2024. That acquisition matters – backed by a company that gets design tools for non-designers. Users have generated 4.5 billion images and 27 million videos since 2022.
The differentiator? Reddit users prefer Leonardo’s web interface over Midjourney’s Discord requirement. Midjourney makes you work inside a chat app, sharing prompts with everyone unless you pay $60/month for stealth mode. Leonardo gives you a private workspace from day one (though free tier generations are public by default).
The Feature Set No One Explains Clearly
Leonardo offers proprietary models (Lucid Origin, Lucid Realism) plus third-party models (Veo 3.1, Sora 2, Kling O1). Each model has a different personality – photorealistic, anime, cinematic.
Model choice? First trap.
Pro tip:Every model shows its native training dimensions – Dreamweaver 7 at 640×832, Leonardo Diffusion XL at 1024×768. Stick close to these ratios or you’ll get bizarre multi-headed figures. Community users call this the “head upon head with an odd leg” problem. 100% preventable if you check dimensions before generating.
Core features:
- Image Generation – text-to-image
- AI Canvas – in-app editing for tweaking generated images
- Realtime Canvas – sketch rough, watch it transform live
- Motion – turn static images into short animated clips (token-hungry)
- Custom Model Training – upload 10-20 images of your style, train a model
That last one? Many pros call custom training the killer feature for brand work. Train once. Generate infinite variations that match your visual identity.
The Actual Workflow: From Zero to First Usable Image
How I do it now, after the token-wasting phase.
Step 1: Sign Up
Go to leonardo.ai. Click “Create an account.”
Google login or email. 150 free tokens, reset every 24 hours. No credit card.
Step 2: Pick Your Model
Image Generation page. Top-left dropdown: Select Model.
First run? Choose Phoenix 1.0 or Lucid Realism. Phoenix produces sharp, photorealistic images. Avoid third-party models (non-Leonardo logo) until you understand token costs.
Step 3: Set Dimensions to Match the Model
Click Input Dimensions. Ratio selector appears.
Shortcut: model trained at 1024×768? Use that exact size or swap it (768×1024 for portrait). Custom dimensions = surreal body horror.
Step 4: Write Your Prompt
Prompt box. Type something concrete: “A red coffee mug on a wooden table, morning light, product photography”. Leonardo auto-enhances basic prompts. Start clear and specific.
Skip negative prompts for now. Add those later once you see what the model defaults to.
(Actually, try generating that coffee mug. The morning light hits different when you realize you just made product photography in 30 seconds that would’ve cost $500 on Fiverr three years ago.)
Step 5: Check Token Cost Before Generating
Bottom-right corner: token cost per image.
Standard 768×768 generation? About 1 token. Enable Alchemy (high-fidelity mode)? Jumps to 5-10 tokens. Add upscaling? More tokens. A single 8-second Veo 3 video costs over 1,000 tokens.
First test: leave Alchemy OFF. You’ll see the difference later. Basic generation quality? Already solid.
Step 6: Generate and Wait
Generation takes 15-90 seconds depending on mode. Progress bar appears.
Done? Four image variations show up (default setting).
Step 7: Download or Refine
Click an image. Download icon (top-right). Done.
Usable AI-generated image. Don’t love it? Tweak your prompt and regenerate. Each run costs another token – iterate with purpose.
The Three Token Traps Nobody Warns You About
Real usage diverges from tutorials here.
Trap #1: Third-Party Models Burn Tokens Without Relaxed Generation Backup
Third-party models (Veo 3, Sora 2, Flux variants, Nano Banana) can’t use Relaxed Generation. What’s Relaxed Generation? On paid Unlimited plans, after fast tokens run out, you keep generating at lower priority for free. Only works with Leonardo’s native models.
You fall in love with Veo 3 video. Burn through monthly tokens by day 10. Now what? Buy top-ups or wait for reset. Community users note many standout models require payment within Leonardo even though they’re cheaper elsewhere.
Trap #2: Paid Plan Tokens Reset on Your Anniversary, Not the Calendar Month
Start a paid plan on July 15? Tokens reset on the 15th of every month, not the 1st. Trips up people comparing usage across team members or budgeting.
Trap #3: Cancel Your Plan, Lose Your Rollover Bank Instantly
Paid users bank up to 3 months of unused tokens in a Rollover Bank. Nice feature.
Cancel your subscription? Every banked token disappears immediately. No grace period. No “use them within 30 days.” Gone.
Thinking about pausing your subscription? Burn through your Rollover Bank first.
When to Upgrade (And When Free Is Enough)
Free tier (150 tokens/day) is generous for hobbyists and casual creators. Most people never need to upgrade.
You hit the wall fast if you:
- Generate 500+ images per day (commercial scale)
- Need private generations (hidden from public gallery)
- Want priority speed during peak hours
- Use video or high-res upscaling regularly
Paid tiers: Apprentice $10/month (8,500 tokens), Artisan $24/month (28,000 tokens), Maestro $48/month (60,000 tokens). Apprentice is the sweet spot – private generations, priority speed, enough tokens for daily use.
Weird thing: even the free tier includes commercial usage rights. You own what you make. But Leonardo retains rights to use, reproduce, and distribute free-tier images. Creating client work? Go paid for full control.
What to Do Next
Open Leonardo right now. Generate one image using Phoenix 1.0 with default settings. Just one. See how it feels. Check your token balance after.
Then try the same prompt with Alchemy enabled. Compare token cost versus quality gain. That’s the moment you’ll know whether you’re a free-tier user or you need to upgrade.
Serious about brand consistency? Upload 10-20 images to train a custom model – generate a hundred variations that all match your style. Leonardo stops being a tool. Starts being infrastructure.
Common Questions About Leonardo AI
Can I use Leonardo AI completely free forever?
Yes. 150 tokens daily, reset every 24 hours. Roughly 150 standard images per day if you stick to basic settings.
Unused tokens don’t roll over to the next day. For casual use? More than enough.
The catch: free-tier images are public by default, Leonardo retains usage rights. Need privacy or full commercial control? Paid plan.
Why do some models cost way more tokens than others?
Third-party models (non-Leonardo logo) can’t use Relaxed Generation after you run out of tokens. Models like Veo 3, Sora 2, Flux variants are computationally expensive and licensed from other companies. Premium features like Alchemy (high-fidelity pipeline) and higher resolutions cost more tokens. Basic 768×768 image: 1 token. Alchemy-enhanced 1024×1024 generation: 10+ tokens. Always check the token cost in the bottom-right corner before hitting Generate. Only way to avoid burning through your daily budget on one image.
What happens if I generate an image using the wrong dimensions for a model?
Every Leonardo model shows its native training dimensions (640×832 for Dreamweaver 7, 1024×768 for Leonardo Diffusion XL). Straying far from these ratios produces distorted renders. Users report “head upon head with an odd leg thrown in.”
The AI tries to fill the canvas but loses coherence when dimensions don’t match training data. Turns out most tutorials never mention this – you only learn it after generating a three-headed portrait and wondering what went wrong. Stick close to recommended ratios or swap width/height (640×832 becomes 832×640 for landscape). One of the most common beginner mistakes. Completely avoidable by checking model specs before generating.