Building the Pipeline
I created a complete end-to-end workflow using LTX 2.3 as the core image-to-video engine. First I generate a carefully written still prompt that captures the atmosphere, camera angle, lighting and action of a chosen Matrix-inspired moment without copying any protected dialogue or exact character likeness. That still becomes the starting frame for LTX 2.3, which I then drive with a short 0–15 second motion script that describes camera movement, character action and environmental effects. The entire process is designed as a practical exercise in cinematic storytelling: learning how to break a film into individual scenes, control pacing, and maintain visual consistency across shots.
The 25 Scenes
Using this method I produced 25 distinct fan-made scenes that follow the narrative arc of the original film while remaining original works. The list includes:
Wake Up Neo
Trinity Hotel Escape
Club Meeting
Office Raid
Interrogation Bug
Red Pill Blue Pill
Pod Awakening
Nebuchadnezzar Rescue
Welcome to the Real World
The Construct
Jump Program
Sparring Program
Woman in the Red Dress
There Is No Spoon
Oracle Visit
Cypher Betrayal Dinner
Lobby Shootout
Helicopter Rescue
Subway Fight
Neo Is Shot
Trinity Kiss Resurrection
Bullet Time Dodge
Seeing the Code
Final Flight
Closing Phone Call
Each scene was generated independently, then reviewed for continuity of style, lighting and mood so the collection can be viewed as a coherent short film exercise.
Learning by Doing
The real goal of this project is educational. By reconstructing a feature-length story scene by scene with LTX 2.3, I practice the same decisions a director and editor face: how long a shot should last, when to change camera angle, how to build tension through movement, and how to keep visual language consistent across an entire sequence. Because every image and motion clip is newly generated and stylized, the work stays firmly in the realm of non-commercial fan study. The finished 25-scene reel becomes both a showcase of the LTX 2.3 workflow and a hands-on lesson in the fundamental craft of making a movie.
Description
Empty underground subway station. Neo and Agent Smith face each other on the platform. They fight hand-to-hand with incredible speed and power. A train approaches in the background. Cold fluorescent lighting, industrial atmosphere, highly detailed combat, 8k, masterpiece.
Cinematic dynamic fight coverage, smooth movement, 8k. Subway station.
0-3s: Neo and Agent Smith stand opposite each other on the platform.
3-6s: They explode into brutal hand-to-hand combat.
6-9s: They exchange powerful blows, using the environment.
9-12s: A train approaches at high speed.
12-15s: The fight continues as the train rushes past. Intense, grounded action, masterpiece quality.
FAQ
Comments (3)
So... 27 years passed, according to "some" we did reached "singularity" and this is what iconic "Matrix" looks like using "Ai".
No disrespect to author, just bothers me that my fav. franchises are butchered this badly. Sometimes I want to unseen things.
Just think, this was done with prompts and a 16GB card and an LTX2.3 model. I'm sure with the same hardware and a couple of LoRas, it could be improved.
I'll definitely review it in a few months... I hope for better.
@MisterMR It's not about You, when You hear news like this one: https://youtu.be/eMf1hxkNahc and then see this (no disrespect to You) You do start to wonder, and ponder, the question, that was already asked, "is this it? is that what RAM / GPU and in the feature CPU had to die for..." ???
PC, personal computing, is no longer personal is now in data centre for a prise.
Neural nets are old tech and it took over 50-60 years to get us to this point. Same goes for RayTraicing... and we can barely do this on top of the line consumer grade GPUs.
This Ai is very similar to VR and from what I see here and when doing my own evaluation of this tek (training models | LORAs) I see story of VR all over again.
And yes models do improve but I personally don't believe that neural network is the way to go. It was computational expensive 20+ years ago what I initially heard about it and it is expensive (both computationally and $$$) today.
And mainly my jab was about this "singularity" thing that OpenAi spits out. Even Nobel prize winner Geoffrey Hinton talks about "singularity" in context of felony that these companies commit. There is no "Ai" that can jump air gapped system (physically unplugged cord from networking equipment) with they all should do if we do believe them that "Ai" is more dangerous than nuclear weapons.
Ok,.. I ranted enough for today :)