Making AI short videos—I’ve stepped into 5 traps

Today I’ll tell you everything, hoping you can avoid some detours.
────────────────
First trap: Learning like crazy right away
When I first started, I studied tools every day, researched techniques, and researched how others were doing it. I was so into studying that I was excited—but when I looked back, I hadn’t finished a single video.
The right order is: make it first, then learn while doing. Once you’ve finished your first one, you understand the process. After you’ve finished a few, you start to feel it. That’s how you improve.
One-sentence summary: Finish first, then perfect
────────────────
Second trap: Trying to save money and credits
Generating images costs money; generating videos costs money—it's painful for sure.
When I first started, I also tried those cheap solutions. The results looked about the same. But after actually using them, all kinds of problems showed up. I spent a lot of time tweaking, and in the end I still gave up.
Compare cost with results, not with price. The cost of making a video is already not high. Saving that little bit of money means you lose time and opportunities.
One-sentence summary: Use the best tools to make the best work
────────────────
Third trap: Being too obsessive about visuals
When I first started, I always wanted the visuals to be super refined. Then I found it was basically impossible—no matter how good the prompts were, there’s still a limit.
Later I realized: many videos with high views don’t have beautifully polished visuals, but the content is captivating and the pacing is right—and they still go viral.
One-sentence summary: Don’t obsess over visuals—good enough is enough.
────────────────
Fourth trap: Forgetting the content itself
This is the biggest trap.
When I first started, I spent a huge amount of time researching tools and techniques, and ended up forgetting that I’m making content.
We’re making videos. The first thing is whether the content is good—not whether the special effects look pretty.
Once I put my attention back on the content, the results changed immediately.
One-sentence summary: We make content, not study AI
────────────────
Fifth trap: Doing too little
If making one try doesn’t feel good, you start making big changes and heavy optimization. But it’s very possible that users just aren’t interested in that direction—switch to another direction instead.
As long as you have enough volume, there will always be one that works.
One-sentence summary: Just do it—when you have volume, miracles happen
════════════════
What tools I use now
You don’t need too many tools. My core right now is Creao
🔹 Video generation—supports Google Veo and Seedance; prompts turn directly into video
🔹 Image generation—up to 4K; generate in real time within the chat
🔹 Voice dubbing—20+ kinds of voices; outputs audio files directly
🔹 Agent automation—once the workflow is set up, you can run it on a schedule; no need to do it manually each time
You can try:
Final line: Don’t just save it—only by doing can you make progress