NEXUS V1 — Archetype-Trained Wan2.1 (1.3B) Finetune

The Most Advanced Wan 2.1 Finetune Ever Created • Powered by the Archetype Training Method™

Nexus V1 represents a breakthrough in open-source AI video generation — a next-generation Wan2.1 model (1.3B) fine-tuned using the Archetype Training Method™, a revolutionary training system that unlocks unprecedented motion consistency, camera control, and scene-level understanding.

This is Version 1 of the Nexus family — a foundation built to rival and surpass commercial models like Runway, Pika, and Kling.

🔥 Trained with the same motion-learning principles that power MDM Nexus — the world’s first self-learning AI video generator.
🔥 Optimized for kinetic realism, character consistency, camera stability, and prompt adherence.
🔥 Built on a unique system of Archetypes that teach the model “how video works,” not just how it looks.

✨ What Makes Nexus V1 Different?

1️⃣ The Archetype Training Method™ (ATM)

Nexus V1 was created with a brand new training philosophy:

Instead of random internet footage or loosely-labeled clips, Nexus V1 was trained on pure structured synthetic datasets using the Nexus system, broken into discrete Archetypes:

Core Archetypes

LINEAR_FORWARD – forward motion, walking, dolly-ins
LINEAR_BACKWARDS – reverse walking & pullback shots
ROTATIONAL 360 – full-body rotation, perfect for stable 360° turns
VERTICAL_MOTION – jumps, elevation changes, lifts
CAMERA PAN / CAMERA TILT / CAMERA TRACKING – pure camera movement training
OBJECT INTERACTION – props, tools, hand-to-object physics
ACTION BEHAVIOUR – punches, dodges, swings, kicks

Each Archetype trains a “skill” that the model can use during generation.
This is why Nexus V1 feels more stable, more grounded, and more cinematic than standard Wan2.

Usage:

CFG: 1.5

Steps: 6+

The model is trained for fast generations at only 6 steps. But the optimal range is 6 to 30 for more compute for complicated shots.

🎥 Camera Control System — Unique to Nexus V1

Nexus V1 supports a full cinematic camera lexicon.
Use these phrases EXACTLY for best results:

CAMERA MOTION

Zoom in
Zoom out
pan left / right
tilt up / down
tracking shot
orbiting camera
handheld shake
steadycam movement
locked tripod static camera

CAMERA TYPES (so far)

DSLR Camera
Phone Camera

CINEMATIC SHOT TYPES

wide establishing shot
medium shot
tight close-up
extreme close-up
overhead shot
over-the-shoulder shot
cinematic depth-of-field
rack focus

ADVANCED CONTROL (exclusive to Nexus)

“camera stays locked to subject”
“camera rotates around subject 360 degrees”
“subject turns showing front, side and back views”
“maintain axis alignment and rotational stability”

These were learned during the ROTATIONAL Archetype training and are unique to Nexus V1.

⚡ Kinetic Action Phrases (Nexus Motion Language™)

These trigger the motion systems trained into the model:

full 360-degree rotation
turns to face the camera
steps backwards while maintaining balance
smooth walking cycle
precise limb coordination
synchronized arm swing
rotational precision around body axis
clean weight shift from left to right leg
dynamic punch with shoulder rotation
high kick with balanced recovery stance

For tools, trades, and object interactions:

hammer strikes nail with correct recoil
carpenter swings hammer repeatedly
boxer throws a jab-cross combination
musician strums guitar with rhythmic motion
chef chops vegetables with controlled wrist action

These were developed through your MotionCraft-inspired training philosophy.

🌀 Prompting Style for Best Results

[TRIGGER WORD], [STYLE (cartoon, photorealistic film etc) SUBJECT (if there are visible subjects then their race, age & gender to combat model bias. And general appearance of subjects such as clothing & hair, textures. if visible, focus on hands & limbs / style, textures, surfaces, yaw, pitch, roll.) We also need ACTIONS (of subjects and environment, if any). LOCATION (foreground & background details). CAMERA AESTHETICS (pan, tilt, zooms etc. foreground lens effects & lighting.) Just a single paragraph to describe the sequence. Do not include any tags or special characters. do not describe the layout of separate images.

1. Put the MOST IMPORTANT motion phrase first

Nexus V1 prioritizes the opening phrase.

Good:
a man performs a full 360-degree rotation, cinematic medium shot, warm lighting...

Bad:
a man standing... and later rotates 360 degrees...

2. Be explicit with physical detail

The model responds exceptionally well to descriptions of:

limb coordination
front/side/back visibility
axis alignment
shoulder rotation
torso orientation

These principles were built into the Archetype system.

3. Keep camera instructions separate from motion instructions

Correct:
a woman performs a slow walking cycle, camera orbiting 360 degrees around her

Incorrect:
a woman walking rotating orbit camera spin around her movement

4. Trigger Anime style (More to come)

ANIME, 2D Animation, flat animation style, hand drawn,

🧩 Example Prompts

High-Quality Cinematic Motion

a ballerina performing a full 360-degree pirouette, showing her front, side and back as she turns, camera locked on tripod, subject rotating in place, soft studio lighting, 4k cinematic realism

Action / Combat

a martial artist delivers a fast jab-cross combination, shoulder rotation clearly visible, precise limb coordination, smooth stance transition, handheld camera movement, cinematic lighting

Walking Stability (ATM Archetype Test)

a man walking forward with perfect smooth motion, synchronized arm swing, stable body axis, slow tracking camera moving backwards, cinematic street environment

Camera Orbit Test

a woman standing still while the camera rotates 360 degrees around her, consistent lighting, clean frame transitions, high motion coherence

💎 Model Strengths

Nexus V1 is extremely strong at:

✔ Motion Coherence & Stability
✔ 360° rotation & turning shots
✔ Walking, running, and body mechanics
✔ Action scenes & kinetic realism
✔ Camera motion control
✔ Object interactions
✔ Prompt adherence
✔ Cinematic shot structure

It produces some of the cleanest rotational movement Wan has ever achieved, thanks to ATM.

⚠️ Known Limitations (This is still only a small model)

We recommend being upfront on CivitAI to build trust.

Occasionally struggles with multi-subject interactions (V2 is addressing this).
Very complex gestures (e.g., multi-spin kicks) may require two-sentence prompts.
Some background drifting can appear during very fast camera motion.
High chaos prompts can break motion stability.

📦 Included Features

Fully trained Nexus V1 checkpoint
Camera Control Guide
Nexus Motion Language List
Archetype Motion Prompt Pack
Example prompt library (starter guide)

🌌 Credits & Acknowledgements

This model is the first public demonstration of the Archetype Training Method™, developed by Anthony Mark (Mark Digital Media) as part of the wider MDM Nexus Project — the world’s first self-learning AI video generator.

Nexus V1 is released to help the open-source community push AI video forward, unlock new levels of motion quality, and inspire the next generation of model creators.

🚀 Next Step

NEXUS V1 isn’t just a finetune.
It’s the new standard for structured AI motion learning.
This is Wan2.1 operating at its full potential — for the first time.

If you're building kinetic datasets, cinematic videos, or experimenting with camera-controlled AI scenes, Nexus V1 is the strongest foundation you can start with.

🆘Help & Support

Join Our Discord!

For the latest updates, news and support for the model, join our discord server to keep updated.
https://discord.gg/UvuGPfRq

Files with SHA256: f970b3bc30aac8a687aedb51e9675f0f0eff0d24cf4921317dde900ebec66236