Welcome to the authentic world of 1920s silent cinema! 🎬

This LoRA is specifically trained for MiniMax H3 to transform your generations into authentic, gritty, and atmospheric vintage film footage. Unlike simple "black and white" filters, this model understands physical film degradation: emulsion damage, gate weave, dust, and chemical staining.

Created for The Civitai H3 Training Contest (Style Category).

⚠️ CRITICAL: HOW TO USE THIS LORA ⚠️

This LoRA is not a magic switch. It learns the connection between specific words and visual artifacts. If you do not prompt for the degradation, the LoRA will not work properly, and you will just get a clean, modern video.

You MUST include artifact keywords in your prompt to activate the style!

🔑 Trigger Word: aged_celluloid

📌 Recommended Prompt Structure:

aged_celluloid, [your subject and action], heavy film grain, severe vertical scratches, flickering brightness, dust spots and water stains, jumpy gate weave motion, crushed black shadows, 16fps choppy stuttering, high contrast black and white.

🚫 Negative Prompt (Recommended):

modern, clean, digital, smooth motion, 4k, 60fps, color, cgi, pristine, restored.

📦 ABOUT THE 3 VERSIONS:

I have trained and included 3 distinct versions of this LoRA, each with nuanced differences in artifact intensity. Choose the one that fits your scene:

Version 1: "Classic Archive" (Balanced) epoch 1

The gold standard. Moderate film grain, natural flickering, and light dust. Perfect for general scenes, portraits, and landscapes where you want an authentic but watchable vintage look.

Version 2: "Subtle Wear" (Light) epoch 3

A gently aged look. Minimal scratches, soft grain, and slight exposure shifts. Best for scenes where you want the feel of old film without distracting from the main subject's details.

Version 3: "Heavy Degradation" (Extreme) epoch 5

Pushes the limits of film decay. Features aggressive vertical scratches, severe emulsion damage, intense light leaks, and violent gate weave. Ideal for horror vibes, chaotic scenes, or "found footage" aesthetics.

🛠️ Training Details:

Base Model: MiniMax H3

Dataset: 114 carefully curated, unrestored vintage film clips focusing on physical artifacts.

Resolution: 512x768 / 832x480

Rank/Alpha: 32 / 16 (Optimized for style retention without overfitting).

🎥 Check out the gallery videos to see the LoRA in action!

If you create something amazing, please share it in the images section and tag me. I'd love to see what you make!