How We’re Using AI to Hack Social Media Videos (and Keep Our Sanity)
Let’s be real. Social media is hungry. Like, “feed me 3 fresh videos before lunch”, hungry. And unless we want to clone ourselves (working on it), we need a way to make cool videos faster without turning into zombies.
Enter: AI.
Case Study (JumpFromPaper)
Why AI? Because We Like Sleep.
AI helps us churn out more content in less time, but here’s the kicker—it still looks good. We’re not talking about stiff robot videos; we’re talking about fun, stylish, and actually watchable content that makes our bags pop. Plus, it lets us try new styles, weird animations, and storytelling ideas without burning out the design team.
Our Secret Sauce (aka Workflow)
Here’s how the AI magic happens:
- Midjourney – We throw in some prompts, and boom, it spits out four wild ideas. Sometimes it nails it. Sometimes it gives us a model with seven fingers holding our bag. (AI is still learning, okay.)
- Photoshop – Where we play plastic surgeon. We fix lighting, add our products, and make sure the bag looks like it belongs in the scene—not like it was photoshopped by your cousin in 2009.
- Midjourney or Higgsfield – Midjourney is great for artsy loops. Higgsfield? That’s where we go full Michael Bay with dramatic camera moves. Each click burns credits, so yes, we cry a little when we overspend.
- CapCut – This is where the chaos becomes a real video. We add effects, music, and polish until it looks like something you’d actually double-tap.
- Done – And just like that, we’ve got a shiny new video that looks like it took weeks, not hours.
Things We’ve Learned (the hard way)
- JumpFromPaper’s 2D-looking bags confuse AI. Sometimes they look like melted pancakes in 3D. More prompts = less pancake.
- AI still struggles with consistent faces. So don’t get attached to “Model A” because she might look completely different in the next clip.
- Wi-Fi speed directly impacts patience levels. Slow internet = dramatic sighing in the office.
The Big Picture
Midjourney, Higgsfield, Photoshop, and CapCut are like our creative Avengers. Each tool is good on its own, but together they help us crank out videos faster, smarter, and honestly, more fun.
What’s next? Nano-Banana!
Final Takeaway
AI isn’t stealing our jobs, but in a way, it’s saving our sanity. It gives us time to focus on the fun, creative stuff instead of fighting deadlines. And hey, if AI ever learns to make coffee too, we’re officially out of excuses.
Workflow (with example)
STEP 1/5
Prompt a scene/concept ideas in Midjourney, and it will generate 4 results (example below):
STEP 2/5
One photo is selected & upscaled to get higher quality in Midjourney:
(Optional) Animate existing generated photo with Midjourney to get a rough ideas on the video:
STEP 3/5
Using AI function in Adobe Photoshop:
- Remove unnecessary elements in the upscaled photo
- Add/replace our products (bag) into an image
- Mood and lighting is matched with harmonize tool
- Details like shadows and highlights added last
STEP 4/5 (Option 1)
Upload the edited image in Midjourney to animate.
It will generate 4 videos options.
Able to make looping animation.
STEP 4/5 (Option 2)
Upload the edited image in Higgsfield.
Generate specific animation/footages/camera movement.
Uses credits to generate (More credit = Higher quality)
STEP 5/5
Compile the AI footages + generated B-Roll and make new project in Capcut for:
- Video editing &composition
- Special effects (optional)
- Add sound effects & music