Image & Video

Reddit user seeks Midjourney-like local image model with diversity, avoiding Flux plastic

16GB VRAM, 64GB RAM – can local models rival Midjourney's aesthetic?

Deep Dive

A Reddit user, out of the local image generation loop and preoccupied with wan 2.2, seeks a high-quality model that feels stable (unlike SDXL), offers diverse outputs per seed, avoids bias (unlike ZIT), and lacks Flux's plastic look—preferring artistic, non-cookie-cutter results, while acknowledging nothing fully matches Midjourney.

Key Points
  • User runs 16GB VRAM, 64GB RAM—targeting high-end local generation.
  • Wants diverse, artistic outputs instead of 'cookie-cutter' or biased results.
  • Current favorites like SDXL deemed unstable; Flux criticized for plastic look.
  • Post received 500+ comments with recommendations like SD3.5 and fine-tuned checkpoints.

Why It Matters

The search for open-source Midjourney alternatives shows growing demand for privacy-first, customizable AI art tools.