How I Made Orion's Echo

For me this film was never only about the story. It was a proof of concept: could one person, working alone, make a full 12-minute animated sci-fi film with the tools we have now? I think the answer is yes. This is how I got there.

Where the idea came from

The idea started with Dennis E. Taylor's book We Are Legion (We Are Bob). I loved the thought of a machine mind waking up a long way from home and having to work out who it is. My first screenplay drafts date back to 2024.

Here's the setup. Orion's Echo is a Xenarion Von Neumann probe that drifts out of a wormhole with its main AI core dead and its memory wiped. A smaller sub-core algorithm, SUBCORE, restarts the main AI, CORE. It tries to rebuild who he was from the archives, and it builds him a virtual meadow to wake up in. I'll leave the rest for the film.

Starting with AI clips (2025)

In late 2025 I recorded the first dialogue lines with ElevenLabs voices. I also explored the look with generative AI tools such as Runway, Google Veo 3.1, Wan 2.5 and OmniHuman 1.5, which gave me about 34 clips and a set of stills. The ship design, the colour palette and the mood all came from that stage. Looking back, those clips were sketches.

v1: getting a whole film on screen

On 20 October 2026 I got the first complete cut together: 12 minutes 9 seconds with the full soundtrack. It mixed the AI clips, AI avatar clips for the talking scenes, a handful of Blender and Unreal Engine shots, and graphics drawn by code. It was rough, but it was a whole film.

v2: every frame from Blender or Unreal

Then I set myself a rule. Every frame you see would be made in Blender or Unreal Engine, and only the voices would be AI. I kept the story, all 72 lines of dialogue, the music, the sound effects and the ship design, and I replaced every AI picture.

The plan came first: a shot list of 58 renders, 39 of them new. SUBCORE became a swarm of glowing nodes that pulses with its voice. It isn't a person, so it never gets a face.

Built by script, with Claude Code

Nothing was modelled by hand. Claude Code wrote and ran the Python scripts that built the ship and drove Blender and Unreal from the command line. They built the levels, keyframed every camera move, rendered the shots, assembled the edit and mixed the sound. The edit plan lists every shot and every line, so shot lengths are timed from the dialogue. Whisper checked that the lines landed where the plan said they should. My job was to make the calls: what to keep, what to cut and what wasn't working.

CORE's three bodies

CORE took the most attempts. He has to speak, so he needs a mouth that moves with his voice.

First he was a MetaHuman, with lip-sync driven straight from the audio. It just wasn't working for me. So I changed the script: CORE has no body until the meadow, and before that he's a glowing crystal that pulses as he speaks. Next came an alien soldier character from Fab, with a jaw shape added in Blender and driven by the loudness of the voice.

In the end I asked for something more basic, a robot that could visibly talk. The final CORE is a simple white-and-red robot built in Blender from 53 rigid parts. His jaw swings open with his voice, LED bars in his mouth and a light in his chest glow as he speaks, and his eyes blink. ECHO, his copy, is the same robot with teal accents. The simplest version was the one that worked.

Thinking like a director

When I watched the first robot cut, the meadow shots were too close and not dynamic enough. All 19 of them were re-blocked as wide and medium shots, because a robot acts with his body, and every camera move has a reason. There's a crane up out of the grass to find him, a low-angle push on "Let's not die in the dark", and a crane up into the stars at night.

Finishing, and the 4K master

The film opens with my own chrome Creative Bandit Studios logo, and the credits match what's actually in it. My first upload looked soft because it was a small share copy, so I rebuilt a 4K master from the full-quality files. YouTube gives 4K uploads a much higher bitrate, and that helps the 1080p version most people watch too.

What was AI, and what wasn't

• AI: the three voices (ElevenLabs), Claude Code writing the scripts and running the builds, and Whisper for timing checks.

• Not AI: every picture in the film is rendered in Blender or Unreal Engine from 3D models and code. The graphics and titles are drawn by code, and the studio intro is my own.

• Music: Echoes of the Deep, Cosmic Serenade and Cosmic Horizons.

• Assets: "Customized Alien Soldier" by Bugrimov Maksim (Fab), the MC Core Motion animation pack, and sound effects from Celine Woodburn, UnrealSFX and Vadi Sound.

Proof that it can be done

It isn't perfect. The kneel animation could be better, the Xenarion sets are simple and a few older shots never got polished. But that wasn't the point. One person, a story they cared about and a pipeline built from scripts made a full 12-minute film, and every frame of it came from real 3D tools. Next time I'll know where the time goes.

Please give it a watch and tell me what you think:

Alex, Creative Bandit Studios