The VHS LoRA model is designed to capture the nostalgic look and feel of VHS footage, delivering a versatile range of realistic and artistic VHS-inspired outputs. The Realistic versions (v1-v3) have been progressively trained on expanding and refined datasets of genuine VHS footage, making them ideal for generating true-to-life vintage visuals. The Artistic version, on the other hand, is trained on simulated VHS effects and glitch aesthetics, allowing for a more stylized interpretation of the VHS look. Each version has been optimized through different annotation techniques for unique control and detail, ensuring that users can recreate the vintage VHS vibe with accuracy.

Base Models:

Flux, SD3.5, SDXL, Pony

Versions/Variants:

- Realistic v1: Trained on tags, focuses on capturing the raw qualities of VHS footage.

- Realistic v2: Trained on captions, delivering refined results with enhanced context.

- Realistic v3: Trained on both tags and captions, offering the most balanced and nuanced outputs in terms of detail and realism.

- Artistic v1: Trained on simulated VHS filters and glitch effects, perfect for creating stylized, retro-glitch visuals.

Recommended Settings:

- CFG:

- Flux: 2.0 - 3.5

- SD3.5: 4.5 - 7

- SDXL & Pony: 5 - 7

- Steps:

- Flux: 20 - 30

- SD3.5: 20 - 60

- SDXL & Pony: 30 - 60

Additional Information:

This model works well for both txt2img and img2img workflows, allowing users to seamlessly introduce authentic VHS effects. Combining Realistic and Artistic outputs can provide creative variations, from true-to-life retro footage to experimental glitch art.

Note:

This model’s results can vary based on prompt details; experimenting with CFG and step ranges will yield different degrees of fidelity to VHS qualities, particularly in Realistic vs. Artistic modes.