---

IAMCCS_anamorph1x – Cinematic Anamorphic LoRA (v2)

IAMCCS_anamorph1x is a cinematic anamorphic LoRA designed for 2.39:1 film-style imagery, developed for Z-Image / FLUX-based generative workflows.

This model does not aim for technical perfection or ultra-definition clarity.
It is built around a different idea of image quality: cinema as optical language.

Anamorph1x – Sample Outputs

All samples generated at 1024×504, LoRA strength 0.8, Z-Image base.

⚠️ Important Note on Image Quality

This LoRA may not reproduce the standard “clean” sharpness often associated with Z-Image outputs.

This is intentional.

True anamorphic cinema is defined by optical imperfection, including:

oval bokeh
edge distortion and stretching
uneven sharpness across the frame
spatial compression and depth ambiguity

These are not defects to be corrected, but part of the cinematic grammar this model is designed to preserve.

If you are looking for:

ultra-clean digital sharpness
uniform focus
illustration-style clarity

this LoRA is not intended for that use.

🎥 Concept

Contemporary generative models and streaming aesthetics increasingly equate quality with total visibility.
Cinema, historically, works in the opposite direction.

Anamorphic optics introduce limitation, distortion, and selective loss of information.
These characteristics protect ambiguity, atmosphere, and authorial identity.

IAMCCS_anamorph1x embraces this logic, translating anamorphic optical behavior into generative space.

🔹 Key Features

Designed for 2.39:1 anamorphic composition
Optimized for cinematic stills, not illustration
Emphasis on:
- lens-driven spatial behavior
- horizontal composition
- filmic texture over digital cleanliness
Preserves optical imperfection as semantic information

🔹 Recommended Workflow

This model works best when used with:

Image reference input
Vision-language models (e.g. Qwen-VL) to describe and reinforce anamorphic intent

A reference-driven workflow helps anchor generation to:

lens behavior
spatial compression
cinematic framing

rather than relying on prompts alone.

🔹 Usage Guidelines

Base models

Z-Image / FLUX-compatible checkpoints

Aspect ratio

Strongly recommended: 2.39:1

LoRA strength

Suggested range: 0.6 – 0.8
Avoid pushing to 1.0 unless intentionally stylizing deformation

Prompt philosophy

Think like a cinematographer, not a prompt engineer
Describe shots, distance, framing, and lens feeling
Avoid illustration, comic, or hyper-digital keywords

Trigger word

anamorph1x
Works well also with:
oval bokeh, anamorphic composition, distorted anamorphic edges

🔹 Versioning

v2 (current release)

Improved cinematic coherence
Embraces optical imperfection as visual language
Suitable for research and authorial experimentation

v3 (planned)

Improved resolution handling
Trained on custom HD anamorphic material
Real anamorphic optics used, including:
- Iscorama 36
- Kowa
More precise control of deformation without losing filmic texture

🔹 Author & Research Context

IAMCCS – Carmine Cristallo Scalzi
Filmmaker, artist, AI cinema researcher

GitHub: https://github.com/IAMCCS
Patreon (research & theory): https://www.patreon.com/IAMCCS
Website: https://www.carminecristalloscalzi.com

This model is part of an ongoing research project on:

optical realism in generative systems
film grammar under streaming-era aesthetics
AI as a cinematic tool, not an illustration engine

A theoretical paper titled
“Anamorphic Imperfection in Generative Cinema”
is available on zenodo: https://doi.org/10.5281/zenodo.18069441 and in excerpt form on Patreon and on the websites: https://www.carminecristalloscalzi.com https://www.faidenblass.com

🔹 License

Released under the MIT License for research and artistic use.

For commercial usage, redistribution, or collaboration inquiries,
please contact the author.

🔹 Final Note

Cinema has never been about perfect visibility.
It has always been about choosing what to reveal and what to withhold.

This LoRA follows that principle.

If my work helped you, and you’d like to say thanks — grab me a coffee ☕

Downloads last month: 47