I Spent 40 Hours Testing Runway, Midjourney, and Leonardo AI – Here’s the Brutal Truth
I’ve been using AI image and video tools for over a year now, mostly for client projects and personal creative experiments. I’ve burned through countless credits, pulled my hair out over inconsistent results, and finally settled on a workflow that works. But I wanted to do a proper, head-to-head comparison of the three biggest names right now: Runway, Midjourney, and Leonardo AI. So I blocked out a week, spent about 40 hours total testing each one on real-world tasks, and here’s what I found.
| Feature / Aspect | Runway | Midjourney | Leonardo AI |
|---|---|---|---|
| Primary Use | Video generation & editing | High-quality still images | Game assets & iterative design |
| Ease of Use | Moderate (steep learning curve for video) | Easy (Discord-based, but intuitive) | Very easy (web UI, lots of templates) |
| Image Quality | Good (7/10) – sharp but sometimes uncanny | Excellent (9.5/10) – artistic, detailed | Very good (8.5/10) – consistent, stylized |
| Video Capabilities | Best in class (text-to-video, video-to-video, inpainting) | Limited (only via external tools or Discord hacks) | None (still images only) |
| Control & Customization | High (masks, keyframes, camera motion) | Low (prompt-only, no direct editing) | High (image-to-image, inpainting, model training) |
| Speed | Slow (especially for high-res video) | Fast (generates 4 images in ~60 seconds) | Medium (takes 30-90 seconds per generation) |
| Pricing | Free tier (125 credits/month), Pro $15/mo (500 credits) | $10-60/mo depending on plan | Free tier (150 credits/day), Pro $10-24/mo |
| Best For | Filmmakers, video editors, motion designers | Artists, photographers, concept designers | Game developers, asset creators, consistent branding |
| Worst For | People who want quick, perfect results | Anyone needing video or precise control | Photorealistic or highly artistic work |
The Setup
I tested each tool on three specific projects:
- A sci-fi character portrait (still image) – to judge detail, style, and photorealism.
- A 10-second video of a city at night (video) – to see motion quality, coherence, and realism.
- A set of 5 game-ready weapon icons (batch consistency) – to test control, iteration speed, and style uniformity.
I used the free or lowest-paid tiers for each, because that’s what most people start with. Let’s get into the nitty-gritty.
Midjourney – The Artist’s Darling
I started with Midjourney because it’s the most hyped. And honestly, for still images, it’s still king. The first thing I noticed is how effortlessly it creates stunning visuals. My prompt for the sci-fi character was: “cybernetic soldier, neon blue eyes, rain-soaked streets, cinematic lighting, 8K, hyperrealistic”. The result? Four variations, each borderline wallpaper-worthy. The skin texture, the reflections in the eyes, the way the rain droplets caught light – it was all there.
But here’s the catch: Midjourney is a black box. You type a prompt, you get an image. You can’t edit a specific area, you can’t change the pose, you can’t adjust the composition without starting over or using a very specific seed. I spent an hour trying to get the soldier’s arm to point upward instead of downward. I tried remixing, using image prompts, even brute-forcing with different seeds. Nothing worked cleanly. The arm stayed down in 90% of generations.
For the city night video? I couldn’t even try. Midjourney doesn’t do video. You can generate a sequence of images and stitch them together, but that’s a whole separate workflow with tools like Deforum or After Effects. It’s a pain.
For the weapon icons, Midjourney was okay but inconsistent. I asked for five different weapon icons in the same flat vector style. The first one looked like a cartoon. The second was photorealistic. The third had weird perspective. I had to regenerate each one multiple times to get a consistent look. It’s not designed for batch production.
Pricing: I used the $10/month Basic plan. That gives you about 3.3 hours of GPU time, which for me lasted about 200 generations. It’s fine for casual use, but if you’re doing a lot of iteration, you’ll hit the limit fast. The $30/month plan gives you 15 hours, which is more reasonable for professionals.
The bottom line: Midjourney is the best at making one-off, gorgeous images. If you need a cover art, a poster, or a concept piece, it’s your tool. But if you need control, video, or consistency, look elsewhere.
Leonardo AI – The Workhorse
Leonardo AI surprised me. I’d heard it was good for game assets, but I didn’t expect it to be this practical. The web UI is clean, and the first thing you see are templates – character sheets, item icons, environments. It’s clearly built for people who need to produce a lot of content quickly.
For the sci-fi character portrait, I used the “Cinematic” preset and the prompt: “cyberpunk soldier, rain, neon, detailed armor, realistic, 8K”. The result was good – sharp, detailed, but not as artistic as Midjourney. The skin had a slight plastic sheen, and the armor looked a bit too clean. But here’s the thing: I could immediately edit it. I used the inpainting tool to fix the soldier’s hand (it was holding a gun I didn’t want) and changed the background from a generic street to a specific alley. It took 10 minutes. With Midjourney, that would have been an hour of frustration.
The city night video? Nope. Leonardo is still images only. No video at all. That was a dealbreaker for my video project.
But for the weapon icons, Leonardo was a beast. I used the “Elemental” preset and generated five icons in one batch: a sword, a bow, a staff, a dagger, and a shield. All came out in the same flat, stylized art style. The colors were consistent, the line weights matched, and I only had to regenerate the shield because it looked too chunky. Total time: 15 minutes. Midjourney would have taken me an hour, and the results would have been less uniform.
Leonardo also has a feature called “Prompt Generation” that uses AI to expand your short prompt into something more detailed. It’s hit or miss – sometimes it adds useless fluff – but it’s handy when you’re stuck.
Pricing: The free tier gives you 150 credits per day. Each generation costs 1-5 credits depending on resolution and model. That’s enough for about 30-50 images per day. The $10/month Apprentice plan bumps that to 2500 credits per month, and the $24/month Artisan plan gives unlimited generations (with a fair use cap). I’d say it’s the best value for heavy users.
The bottom line: Leonardo is the tool for production work. If you need consistent assets, iterative design, or a lot of control, it’s your best bet. But it can’t touch Midjourney for pure artistry, and it has zero video capabilities.
Runway – The Video Frontier
Runway is a different beast. It’s not just an image generator; it’s a full video editing suite with AI tools. I’ve used it for client projects before, so I knew what to expect, but I wanted to see how it stacked up for these specific tasks.
For the sci-fi character portrait, I used the “Text to Image” feature in Runway’s Gen-2 model. The prompt was the same as before. The result? Okay, but not great. The face was a bit blurry, the lighting was flat, and the cybernetic details were muddled. It looked like a mid-tier AI image from 2023. Runway’s strength is not still images – it’s fine for quick concepts, but you wouldn’t use it for a final render.
The city night video, though? That’s where Runway shines. I used the “Text to Video” feature with the prompt: “nighttime cityscape, rain, neon signs, slow camera pan, cinematic, 4K”. The output was a 4-second clip (the free tier limits to 4 seconds), but it was coherent. The cars moved naturally, the rain fell consistently, and the camera pan was smooth. I then used the “Video to Video” feature to apply a “Film Noir” style, which added a cool blue tint and increased contrast. The whole thing took 5 minutes.
I also tested the inpainting feature on a video. I had a clip of a person walking down a street, and I wanted to replace a billboard in the background. In Runway, you can mask the billboard, type a new prompt (“neon sign saying ‘Runway’”), and it changes it in the video. It’s not perfect – the edges flicker a bit – but it’s miles ahead of anything else I’ve seen.
For the weapon icons, Runway was useless. It’s not designed for batch image generation. You can generate one image at a time, and the quality is mediocre. I wouldn’t use it for that.
Pricing: Runway has a free tier with 125 credits per month (enough for about 25 text-to-video generations). The $15/month Pro plan gives you 500 credits, and the $35/month Unlimited plan is, well, unlimited. But note: high-resolution video costs more credits. A 4-second 4K clip might cost 20 credits, so you burn through them fast.
The bottom line: Runway is the only tool here that does video well. If you’re making short films, ads, or motion graphics, it’s essential. But for still images or batch production, it’s not competitive.
The Real-World Performance Test
I ran each tool through a stress test: generating the same concept – “a futuristic city street at night” – and comparing the results.
Midjourney: Gave me a stunning, atmospheric image. The neon lights reflected off wet pavement, the buildings had intricate architecture, and the overall mood was perfect. But I couldn’t change the camera angle or add a specific element without starting over.
Leonardo: Produced a solid, game-ready image. It was less artistic but more functional. I could easily edit out a trash can I didn’t like and add a flying car. The style was consistent with other images I’d generated.
Runway: For still images, it was the weakest. The city looked generic, and the details were soft. But when I turned the same prompt into a video, it came to life. The rain moved, the lights flickered, and it felt like a real place.
Pros and Cons (No Fluff)
Midjourney
- Pros: Unbeatable image quality, great for artistic projects, fast generation.
- Cons: No video, no direct editing, inconsistent for batch work, expensive for heavy use.
Leonardo AI
- Pros: Excellent control, great for batch production, affordable, good consistency.
- Cons: No video, images can look slightly plastic, not as artistic.
Runway
- Pros: Best video generation, powerful editing tools, good for motion designers.
- Cons: Weak still images, expensive credits, not for batch work.
The Winner (and Why)
If I had to pick one tool to rely on for the next year, it would be Leonardo AI. Here’s why: it’s the most versatile for actual work. I can generate a character concept, iterate on it, produce a full set of assets, and do it all without breaking the bank. The control features – inpainting, image-to-image, model training – are indispensable for real projects. Midjourney makes prettier pictures, but Leonardo makes usable pictures.
That said, I still use Midjourney for hero images and concept art. And I use Runway for any video work. But if you forced me to choose one tool to handle 80% of my needs, it’s Leonardo.
Final Thoughts
The AI image space is moving fast. By the time you read this, all three tools will have new features. But the core strengths are clear: Midjourney for artistry, Leonardo for production, Runway for video. Don’t let the hype fool you – pick the tool that matches your actual workflow, not the one with the flashiest demos.
I’ll probably get hate from Midjourney fans for this, but I’d rather have a tool that lets me fix a mistake in 10 minutes than one that forces me to regenerate 50 times. That’s the difference between a creative toy and a professional tool.
