Performance driven AI videos will dominate 2026 This workflow gives you maximum control: 1. Use this 3x3 dialogue prompt in Nano Banana Pro to create different shots of your scene 2. Animate via Kling AI 2.6 Motion Control to transfer your own acting & lip sync Prompt below:

Videos generated with @Kling_ai 2.6 Motion Control and NB Pro via @freepik Upload the character references, a vehicle reference and a reference that shows the location Prompt: Use the uploaded image(s) as a reference for location, lighting, color grading, mood, style and environment. Use the uploaded characters as a character reference. Keep characters consistent across all panels. Car: [CAR DETAILS] Role: The [CHARACTER A] is the driver. The [Character B] is the passenger. relaxed posture, one arm resting on the door. Build a single 3x3 cinematic storyboard grid, panels clearly separated by thin borders, counted left to right, top to bottom, adhering strictly to the shot structure below: 1) Aerial Establishing Shot (Drone Master) — extreme long shot with negative space: [LOCATION DETAILS] 2) Exterior Windshield Two-Shot (Hood Mount Master): match the reference’s front windshield framing and grade; both characters visible; [CHARACTER A] clearly driving with hands on wheel; [CHARACTER B] passenger leaning back. 3) Interior Medium Single (Driver): medium shot focused on [CHARACTER A] driver with the steering wheel and gauges clearly in front of her; hands steady on wheel; 4) Interior Medium Single (Passenger): medium shot on [CHARACTER B] seated as a passenger. 5) Interior Close-Up (Driver/Reaction): tight close-up on [CHARACTER A]'s face (and hands on the wheel; shallow depth of field.) 6) Interior Close-Up (Passenger/Reaction): Tight facial framing of [CHARACTER B]. Eyes are focused on the road ahead. 7) Rearview Mirror Insert (Eyes): extreme close-up of the rearview mirror reflecting [CHARACTER A] driver’s eyes looking at the road forward. 8) Interior Back Seat Master: wide from back seat looking forward; [CHARACTER A] is visibly driving (hands on wheel), [CHARACTER B] passenger beside; [CAR INTERIOR DETAILS & LOCATION DETAILS] 9) Front/Hood POV (Forward Tracking) Low POV from the front of the vehicle facing forward; the [Vehicle]’s hood/front elements lead the eye; road/path vanishes into the distance [LOCATION DETAILS]

I recorded a few videos with my phone and used the footage to drive the motion and lip sync. You can also prompt specific motion, like background movement, light reflections and more. Next time I’ll try to put together some nice before and after comparisons

cool

Thank you!

This explains exactly my own process, its really well done and gives a great window into the best way to go about narrative performance movies. There will always bd shots within the grid that you dont use, but extract them anyway and modify the stills for later scenes

Exactly! You don't need to squeeze all your panels into the scene just for the sake of it, but it's always useful to have them ready, just in case you need them for something else

I find it better to have two returns in NBP, then you have more scenes to choose from although you will get some inconsistencies in continuity and environment, but cutting, matching and obscuring is a skill in itself Editing for me isn't editing, its the art of hiding things

Holy fucking shit. If AI movies with camera angles like this are popular in 2026, it’s going to be a total fucking disaster. Nobody shoots movies like that! Nobody sets up camera angles like that!

Quick question in your prompt do you come up with these different angles or does the LLm do it, then you create a story around it?

It’s honestly a mix of both. I’m currently using NotebookLM to explore interesting camera angles that fit a scenario I have in mind and a rough script I’ve written and based on that, it recommends angles, when to use them and why From there, I refine the script and sometimes I discover angles I really like and then rewrite the script around those shots. So it’s definitely a combination of both!

Excellent! Thanks for sharing!

This is dope!

Thanks!

acting skills are the new prompt engineering in 2026. using nano banana pro for the base plates is smart because consistency is usually where these workflows die. if you nail the lighting in the 3x3 grid the kling transfer feels way less uncanny. the days of just typing walk forward are over you actually have to perform now to get the good stuff

cool

haha I love this! Another great one brother!

Haha thank you so much brother! Much love!

as always. DOPE work, dude. Thanks for sharing! Love this duet!

Huge thanks Marco! Much love! ❤️

When I watch videos like this that feel real in every sense, it kind of makes you wonder: when AI hits singularity, how real these characters will become in their worlds if their worlds follow the same "observed reality" we live in. Very cool characters! I love them ;-)

Very cool,what did you do for the voices?

Thank you! I'm using Elevenlabs :)

Aha got it,so you did the voices first and then you did the lip sync or did you used that voice2voice feature?

Both! For some shots I used the voice changer feature and for others I generated the voice with TTS and then did the lip sync by performing it myself

Very cool,thanks for sharing!

Nice work. Beautiful women, handsome men, cars, nature—the scenes are beautiful and vivid.

Thank you! Happy you like it!

Excellently done sir!

Thanks a lot!

Super insightful! Using a 3x3 dialogue prompt with motion control seems like a powerful workflow for cinematic AI videos.

@NossaCoisa1

good work

@noromachine Thanks!

video is great but voice doesn't match at all the scenes

¡Trabajo excelente! 👏✨

Excellent editing 😸

Thank you!

Interesting! Kling AI 2.6, you say? 👀 I'm intrigued!

Right, Kling AI 2.6 Motion Control

Great job! But, which is the prompt for animate the 9 images of the grid in Kling 2.6 Motion control?

The voice over animations are total shit

thanks for share it

That's excellent information. Thank you so much!!

Love to see it! Thanks for the tips

Always, happy you like it!

Yaaaasssss

This is the right take. “Performance-driven” is the key because it adds controllability: you’re separating shot generation (3x3 prompt grid) from motion/performance transfer. That’s basically the missing piece for scalable ad testing: same script, same acting, multiple visual variants. Curious what you’re using to keep character + wardrobe consistency across all 9 shots?

What prompt did you use for the image grid? Lovely storyboard!

I created same simulacrum:

We’ve seen it time and time again, tools don’t matter as much as how you chain them together.

It looks fantastic—super smooth and incredibly engaging. Thanks for sharing the prompt and your creative process; I've learned a lot from it.

Impressive..... I didn't know that Kling motion could animate background too

Cool

This is cool