Testing MiniMax Image Consistency: Separate Images Beat Character Sheets
vuse2121 · reddit · 2026-09-14
The author ran tests on MiniMax's image input consistency, feeding five separate shoulder-up angles as individual images. Subject consistency turned out to be drastically better than with a traditional character reference sheet.
A more complex setup — three base angles (straight-on, 45-degree, profile) combined with vertical variations (high, eye-level, low angle) — muddied results once prompt tweaks entered the mix.
The open question: is multi-angle input overcomplicating things, and does it all just come down to base image quality and resolution?
More from Multimodal
- HyperFrames + HeyGen: decent video from a basic prompt in ~10 minutes — toolstelegraph · 2026-10-30
- Hailuo MCP + Claude: chaining motion graphics generation through an agent — aziz4ai · 2026-10-30
- AI short film Nexus ep 7 cost $28,775 to make, creator shares full breakdown — NightsRadiant · 2026-10-30
- LayerShop: open-source Photoshop-style editor with ComfyUI local AI on layers — HonestTicket6013 · 2026-10-30
- AI video prompt goes viral: rally car racing a collapsing volcanic mountainside — umesh_ai · 2026-10-30
- Redditor finishes a short dinosaur movie made with Veo 3 and Seedance — HeadOpen4823 · 2026-10-30