Transform your generations into authentic, hand-painted silent film title cards with visible physical decay! 🎬📜

This Concept LoRA for MiniMax H3 specializes in generating vintage intertitle cards — the iconic text screens used in 1920s silent cinema. Unlike modern digital fonts, this model learns the physical imperfections of real celluloid title cards: cracked paint, uneven ink density, emulsion damage, water stains, and the characteristic flickering jitter of old film projection.

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

⚠️ CRITICAL: HOW TO USE THIS LORA ⚠️

This LoRA is NOT a magic switch. It learns the connection between specific artifact keywords and visual damage. If you do not explicitly describe the film degradation in your prompt, the LoRA will produce weak or invisible results.

🔑 Trigger Word: aged_intertitle

📌 You MUST include artifact keywords such as:

cracked paint texture

chipped letter edges

uneven ink density

heavy film grain

vertical scratches cutting through letters

flickering brightness

jumpy gate weave

water stains

emulsion damage

🚫 Without these keywords, the trigger word alone will NOT produce the full vintage effect!

📦 TWO EPOCH VERSIONS INCLUDED:

Through extensive testing, I discovered that different training epochs produce distinctly different artifact profiles. Both versions are included so you can choose or alternate between them for maximum variety:

Version 1: Epoch 7 — "The Oxidizer" (Rust & Decay Focus)

Emphasizes chemical staining, rust-like emulsion damage, and edge decay. Letters appear more weathered, with prominent circular water stains and faded corners. Ideal for eerie, atmospheric, or dramatic title cards where you want the text to feel old and damaged.

Version 2: Epoch 10 — "The Projector" (Shake & Scratches Focus)

Emphasizes intense gate weave shaking, heavy vertical film scratches, and aggressive flickering. Letters vibrate more violently, simulating a damaged film strip running through a broken projector. Ideal for chaotic, fast-paced, or "found footage" aesthetics.

💡 Pro Tip: For the best results in a full silent film sequence, alternate between Epoch 7 and Epoch 10 for different title cards. This creates visual variety and prevents the audience from noticing repetitive artifact patterns. You can also stack both at reduced weights (e.g., 0.5 + 0.5) to combine their effects!

📌 RECOMMENDED PROMPT STRUCTURE:

aged_intertitle, full screen title card, the text clearly reads "[YOUR TEXT HERE]", [describe font style: ornate art deco border / simple serif letters / decorative frame], white text with [cracked paint texture / uneven ink density / chipped edges], heavy film grain on the black background, [deep vertical scratches / circular water stains / emulsion damage], [flickering brightness / jumpy gate weave / unstable jittery motion], 16fps choppy motion, high contrast black and white.

Example Prompt:

aged_intertitle, full screen title card, the text clearly reads "THE MYSTERY DEEPENS", ornate art deco border with chipped and faded paint corners, white text with uneven ink density and cracked paint texture, heavy film grain on the black background, deep vertical scratches cutting through the letters, slight flickering brightness, jumpy gate weave making the text vibrate, 16fps choppy motion.

🚫 Negative Prompt (Recommended):

modern, clean, digital font, smooth motion, 4k, 60fps, color, cgi, pristine, restored, perfect text, sans-serif, minimalist.

🛠️ Training Details:

Base Model: MiniMax H3

Dataset: 100+ carefully curated vintage intertitle cards from unrestored 1920s silent films, focusing on physical text damage and film degradation.

Resolution: 512x768

Rank/Alpha: 48 / 24 (Optimized for text structure retention and complex artifact learning).

Learning Rate: 3e-5 (Conservative to prevent over-smoothing of text edges).

Epochs: 7 and 10 (Both versions included for different artifact profiles).

🎥 Check out the gallery videos to see both epochs 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!