The last few weeks have been busy in a quieter way.

There actually was a big release just a few days ago. Kling 3.0 dropped, and I’m in the middle of digging into it now, trying to understand where it really shines and where it doesn’t. I’ll do a proper breakdown once I’ve spent enough time with it.

But leading up to that release, things had been relatively calm on the surface. And that gave me space to focus less on chasing what’s new and more on actually finishing one project and setting up the next one properly.

That slower stretch is where a lot of the real work happened for me. Wrapping a film that was almost there, figuring out why it still felt unfinished, and then taking those lessons straight into the setup of the next project.

What I wanted to share here are a few of the tools and AI-driven workflows I leaned on during that phase. Not theory. Not hype. Just the things that helped me get one project across the finish line and set the next one up so it doesn’t fall apart halfway through.

On my last project, the visuals were already working, but the film still felt flat. The missing piece wasn’t better prompts or better shots. It was sound.

Not music. Environmental sound.

Here’s the simple process I used.

1. Generate or search sound effects with AI
I used ElevenLabs for sound effects. Sometimes I searched their existing library for basics like footsteps or room tone. Other times I generated custom sounds with short prompts like:
“The sound of two people walking down a street with light city ambience.”

2. Download and organize everything
Once I found sounds that worked, I downloaded them into a dedicated audio folder inside the project so I could reuse them later.

3. Manually place sounds in the timeline
I dropped the sounds directly into the edit and lined them up visually and by ear. Footsteps, basketball bounces, background ambience. Nothing fancy, just intentional placement.

Some sounds didn’t work perfectly. Basketball dunks were rough. For those, I pulled sound effects straight from YouTube.

Not elegant. But the film got finished.

After sound design, the film felt better, but still a little too clean. Too digital.

I used AI to help solve that as well.

1. Ask AI to analyze the film
I dropped the video into Gemini and asked what could be improved. One thing it pointed out was inconsistent texture across scenes.

2. Generate a LUT with ChatGPT
I asked ChatGPT what file types editors use for color and overlays. That led me to cube files. Then I asked it to generate a 16mm-style film grain LUT. It asked a few clarifying questions about intensity and overall feel.

3. Apply and dial it in inside the editor
I imported the cube file into CapCut, added an adjustment layer, and adjusted the intensity by feel. Light grain early in the film. Slightly heavier later so everything felt cohesive.

Same footage. Same story. Completely different feeling.

While finishing that film, I was also starting the next one. This is where most AI projects either stay cohesive or slowly drift apart.

Here’s the high-level setup I’m using now.

1. Lock the story first
If there isn’t a script, I use ChatGPT to get one on the page, then Gemini to analyze structure and gaps.

2. Create a single source of truth
The script and assets live in a shared Drive folder that both ChatGPT and Gemini reference. No guessing. No relying on memory.

3. Build a character and scene bible
Before generating anything heavy, I extract characters, wardrobe, props, locations, and recurring details into one document. Once characters are chosen, they’re locked and treated like casted actors.

That entire setup process is documented step by step in a short guide based on how I’m actually working right now. Link here.

You don’t need to follow every step.
You don’t need to change everything at once.

But even borrowing a few of these ideas can make finishing your current project easier and starting the next one a lot smoother.

That’s really been the theme lately. Less chasing. More finishing. More intention up front so the creative part stays fun instead of frustrating.

More soon.

Khalil

If you want to go deeper, my guides are available on my Stan store.

AI keeps coming up at work, but you still don't get it?

That's exactly why 1M+ professionals working at Google, Meta, and OpenAI read Superhuman AI daily.

Here's what you get:

Join 1M+ pros