Ever dreamt of directing your own fantasy film? Not long ago, this idea would’ve stayed just that—a dream. But with the rapid progress of AI, more and more “text-to-video” generators are emerging, turning your prompts into impressive videos in just minutes. While what AI can achieve is amazing, it’s far from perfect. In this article, we’ll take a closer look at three tools—KLING AI, Runway ML, and LUMA—and explore the challenges and possibilities that come with this new way of creating content.
I was inspired to write this after seeing the creative videos produced by the KLING AI community. https://klingai.com/community/skit
Some of the results are truly professional—whether they’re realistic or surreal, with surprising twists or clever punchlines, they all have the ingredients to go viral.
What are AI video generators?
How to work with an AI video generator
At first glance, working with an AI video generator seems straightforward—just like using ChatGPT. You input a prompt and wait a few minutes for the result. But the outcome can sometimes disappoint, especially if you’ve paid for credits and the result doesn’t meet your expectations of a perfect viral clip.
I couldn’t resist trying out video creation with prompts myself. The results were mixed, so I’ve gathered a few tips to help you avoid disappointment.
My approach
I decided to test three different AI video generators. I tried:
- Creating a video from my own photos
- Creating a video using AI-generated images (Midjourney, FLUX, Mistic, Adobe Firefly)
I also experimented with the Cassovia Code logo, which you can see here.
For each AI generator, I used the same prompt and the same set of photos or images. In some cases, I had to tweak the prompts. Through trial and error, I found that it’s better to avoid multi-step instructions. Transitions between photos didn’t always work as expected.
Since the final videos were only 5 seconds long and transitions were occasionally jerky, I had to edit the videos, add music, and merge the clips using an editing tool like CapCut.
Key tips for using AI video generators:
- Create a script or pick a theme. Keep it simple with minimal action.
- Prepare your own photos or AI-generated images. (Purely text-based video generation often produced weak results for me).
- Refine your prompts, keeping in mind that most generators struggle with multi-step instructions.
- Describe camera movements. The camera can be static, smoothly move (panning, tilting), or follow an object (tracking), with options for zooming and rotating (orbiting). You can switch perspectives between wide shots, close-ups, or bird’s-eye views.
- Check out each tool’s documentation or YouTube tutorials to better understand prompt terminology.
Common mistakes to avoid:
- Overly complex scripts — AI struggles with processing complicated instructions. It’s better to break the script into smaller, more manageable steps.
- Vague or unclear prompts — Ambiguous instructions can lead to disappointing results. Be as specific as possible about what you want to achieve.
- Using the wrong tool— Different AI tools have different strengths. Some are better for realistic animations, while others excel at abstract visualizations.
- Low-quality or unsuitable images —If you’re using photos, make sure they’re high quality with good lighting. Poor-quality or unsuitable images can lead to distorted figures or faces (like the runners in our video).
- Not checking the process — AI can sometimes generate content that strays from your original idea. Regularly review and adjust your prompts to stay on track.
Comparison of the tools:
| Tool | Price (5s video) | Generation Speed (5s video) | Video Quality | Special Features |
|---|---|---|---|---|
| KLING AI (KLING 1.5) | €0.49 (promo €0.14) | 4-10 minutes | High fidelity and scene detail | Advanced camera settings and relevance controls |
| Runway ML (Gen-3 Alpha Turbo) | €0.55 | <1 minute | High quality, realistic camera movement | Fast generation, ideal for experimentation |
| LUMA (Dream machine) | €0.14 (€0.20 for commercial use) | 2-3 minutes | Lower resolution, strong in morphing between photos | Free generation options, great transitions |
The future of AI-generated videos
AI video technology is evolving rapidly and has the potential to transform video creation—whether for commercial or artistic purposes. Comparing videos made in 2023 and 2024 reveals significant improvements in quality, realism, and speed.
@sebastienjefferies text to video 1 year later, imagine what it will look like in 2025 🤯 The top video is a famous scene of Will Smith eating spaghetti which has been generated using AI and the bottom scenes are from Open AI’s latest announcement called ‘Sora’ #openai #ai #aitools #aivideo #videographer #filmmaker ♬ Kids - Old Jim & Zombic
@willsmith This is getting out of hand! #aivideo #sora ♬ original sound - Will Smith
As technology advances, we can expect AI systems to handle more complex prompts, generate longer videos with better narrative structures, and even integrate interactive elements. AI videos could become a new standard in digital marketing, but this will place greater emphasis on originality and authenticity.
AI tools will allow creators to produce content faster and more cheaply. Plus, the availability of these tools means video creation won’t just be the realm of professionals—it will be open to anyone with a good idea.
In conclusion, the world of AI-generated videos is just getting started, and the possibilities are nearly limitless.
What to read next? Try 6 Top AI Tools for SAP Consultants and Beyond.
Lukáš Figely
Lukáš "Lucky" Figely is our UX/UI designer who also has a strong grasp of programming, allowing him to merge impressive design with functional code. He is passionate about exploring AI innovations, particularly those that push creativity to new dimensions.
