Replicating the Midjourney Look with Open Source Models: Workflow & Prompting Discussion
Disastrous_Pea529 · reddit · 2026-08-05
A paid subscriber posted asking for a complete open-source workflow to replace Midjourney. They noted that images generated by open-source models often look flat and oversaturated, struggling to achieve MJ's polished, cinematic vibe.
They asked the community for practical advice on several key areas:
- Base Model Selection: Among newer models like Krea 2, Z-Image, or Ideogram, which one gets closest to the MJ look?
- LoRAs & Prompting: Do "MJ style" LoRAs actually work? Can traditional MJ prompt syntax (like comma lists and stylize values) be directly translated?
- Style References & Post-Processing: Is there an open-source equivalent to MJ's sref? And how much of MJ's "polish" relies on upscaling and detail passes versus the model itself?
More from Multimodal
- Next-Gen AI Video: Minimax M3 Generates Surreal Vid2Vid Transformations — Promptmethus · 2026-08-05
- Test Drive: Qwen 3.8 Max Generates a Functional 3D Racing Game — iamfakhrealam · 2026-08-05
- ByteDance Launches SeedRealtime: Free Real-Time Video Chat in Doubao App — xiaohu · 2026-08-05
- Testing MiniMax H3: Stunning Single-Character Results at 85% Lower Cost — dorocreator · 2026-08-05
- MiniMax H3 Local Test: 5-Second Video in 50s on 96GB RTX PRO 6000 — Practical_Low29 · 2026-08-05
- Sand.ai Open-Sources 100B-Param Video Model MAGI-2 at 1/10th Inference Cost — 量子位 · 2026-08-05