For a long time, my creative output was still images. Paintings, digital illustrations, concept art. I'd post them, people would like them, and that was the whole exchange. A flat rectangle of color frozen in time. I was fine with that for years.
Then a friend sent me a video of her painting breathing. Literally, the leaves in the background swayed, the water surface rippled, and the light shifted as if the sun was moving behind clouds. She'd made a 15-second loop from a single digital illustration and it had 80,000 views on Instagram in two days. The still version of the same piece had gotten maybe 3,000 likes.
That was six months ago. I've been testing AI video tools obsessively since then, and I want to share what I've actually learned, not the promotional copy, but what happens when you sit down and try to make your own work move.
Why This Is Different From AI Image Generation
When I first heard "AI video tools," I assumed it was just image generation with time added. It's not. The creative problem is completely different. A still image needs to look right. A video needs to look right and feel physically plausible across time. Objects have to move in ways that make physical sense. Lighting has to shift consistently. A character's hair has to flow in one direction, not teleport between frames.
Getting that right is a harder technical problem, which is why AI video lagged behind AI images by a couple of years. But the tools have caught up enough that I now use them in my actual creative practice, not just for experiments.
The Four Tools I Tested
Runway Gen-3 Alpha: The Professional's Pick
Runway is the tool I reach for when quality matters most. Gen-3 Alpha produces video that holds together, textures stay consistent, motion is smooth, and the outputs look like something that was filmed, not generated. The image-to-video mode is where I spend most of my time: upload one of my digital paintings and describe how you want it to animate.
The workflow that works best for me: I generate a still image in Midjourney first, making sure composition and lighting are exactly right. Then I upload that to Runway and write a motion prompt that's very specific about what moves and what stays still. "The water surface ripples in small waves moving left to right. The trees sway gently. Everything else is still." Specificity matters here, a vague prompt like "make it cinematic" produces unpredictable results.
What Runway does better than anyone: camera movement. You can specify a slow pan, a push-in, a zoom-out, and it executes them with remarkable control. For visual artists creating looping content, this alone is worth the subscription.
The price is real ($15-35/month depending on plan), and credits burn fast if you're generating a lot of 10-second clips. I spend about 4 hours on a finished piece and use maybe 30-40 credits.
Best for: High-quality animation of detailed illustrations, camera movement work, professional output
Pika 2.0: The Fastest Iteration Loop
Pika is where I test ideas before committing to Runway credits. The generation is faster, cheaper, and honestly pretty good for mid-tier work. The interface is the most intuitive of any tool I tested, and the "Pikaffects" system gives you one-click motion options (explode, deflate, melt, grow) that produce genuinely interesting results for abstract and stylized art.
Where Pika shines for artists: character animation. I have a series of portraits I've been creating, people in Accra in various settings, and Pika handles subtle head movements and eye blinks better than Runway. There's something uncanny-valley-adjacent that Runway sometimes produces when animating faces. Pika's face animation feels softer and more natural.
The limitation is consistency across longer clips. Anything over 4-5 seconds and the texture coherence starts to slip. Fine for social media loops. Not for anything longer-form.
Best for: Quick iteration, portrait animation, stylized and abstract work, social media clips
Kling 1.6: The Unexpected Contender
I'll be honest: I came in with low expectations for Kling, the Chinese-developed tool that arrived with a lot of hype and limited access. The hype turned out to be mostly warranted. Physical simulation is where Kling beats everything else I've tested. Cloth, water, fire, smoke, anything that follows real physics: Kling handles it in a way that feels modeled rather than guessed.
I animated a painting of a woman wearing kente cloth, and Kling got the fabric movement right. The pattern stayed consistent (a huge challenge for patterned textiles), and the drape and movement looked like actual woven fabric, not a painted surface that happens to wiggle. I've gotten worse results from hand-keyed animation.
The English interface has improved significantly. Early versions required careful workaround prompting to get the results you wanted. Now it's straightforward.
Best for: Anything involving fabric, fluid dynamics, fire, or realistic physical simulation
Kaiber: The Artist's Loop Machine
Kaiber sits in its own category because it's not trying to be photorealistic or create general-purpose video. It's designed specifically for artists who want to create looping visual content with a distinctly stylized look. The "transform" mode takes your existing video or image and passes it through a visual style, effectively painting it in real-time.
The results look like nothing else in the AI video space. There's a quality that's somewhere between a music video and a moving painting, it's clearly AI-generated, but it's AI-generated in an interesting way rather than a bland way.
I use Kaiber primarily for music video work and for visual content that I want to feel handmade and artistic. It's the opposite of chasing photorealism. If you're a digital artist who makes stylized work, Kaiber's aesthetic fits better than the others.
Best for: Looping art, music video content, stylized animation that should feel artistic rather than realistic
The Workflow I Actually Use
After six months of experimentation, this is the pipeline that produces my best work:
- Create the still in Midjourney or by hand. Get the composition, lighting, and color exactly right before thinking about motion. The motion will emphasize whatever is in the still, so the still needs to be solid.
- Test motion concepts in Pika. Cheap, fast, good enough to evaluate whether an animation idea is working. I generate 5-8 variations here.
- Pick the best concept and execute in Runway or Kling. For fabric-heavy work or anything physical: Kling. For camera movement and detailed environments: Runway.
- Loop it in Premiere or DaVinci Resolve. AI video tools produce clips, not loops. Getting a seamless loop requires a tiny bit of post-processing: crossfading the end back into the beginning. Takes five minutes and completely changes how the content feels.
- For stylized content, take the Runway output into Kaiber. Run it through a style transform to add the handmade quality I like in my work.
A complete animated piece takes me 3-5 hours now. Before these tools existed, I was spending 2-3 days on motion work (I had to hand-key everything in After Effects, which I'm technically okay at but not fast at).
What Still Doesn't Work
I want to be honest about the limitations, because the marketing materials for all of these tools show their best outputs, not their typical ones.
Text in images is still a nightmare. If your illustration includes readable text (signage, book titles, anything with letters) expect the AI to mangle it the moment the camera moves or the surface shifts. I've learned to keep text static or remove it before animating.
Very detailed, complex scenes lose coherence over longer durations. A crowded market scene, a cityscape with many buildings, an intricate pattern: all of these start to blur and shift after 3-4 seconds. The tools work best on images with a clear subject and a relatively simple background.
And loops are still harder than they should be. Getting a truly seamless loop (where the end frame transitions naturally back to the first frame) requires either careful planning of your initial still or post-processing work. None of the tools generate guaranteed seamless loops natively, though Runway's upcoming loop feature looks promising.
What This Means for Visual Artists
Here's the thing I keep coming back to: these tools don't replace any skill I have. They extend what I can do with skills I already have. I still create the original image. I still make every compositional decision. I still choose the mood, the color story, the subject. The AI handles the technical labor of moving that image through time in a way that looks plausible.
From Accra, where I don't have access to a studio full of animators or motion graphics specialists, this matters a lot. My work can compete visually with studios that have resources I'll never have. Not because AI is magic, but because it gives a solo creator tools that used to require a team.
If you make still images and you haven't tried animating them yet, I'd start with Pika. Create a free account, upload your best piece, write a specific motion prompt, and see what happens. The first time one of your paintings breathes, you'll understand what I'm talking about.