Arabic text-to-image shootout: GPT Image flubs Arabic while rivals nail it
aziz4ai · x · 2026-11-11
aziz4ai ran eight identical prompts across MiniMax H3, Seedream 5.5, Nano Banana 2.1, and GPT Image 2.6 in a five-style Arabic image-gen shootout (risograph posters, claymation dioramas, anime scenes, baroque oil paintings). Surprise finding: on the Arabic headline poster, Nano Banana, MiniMax and Seedream rendered the text correctly while GPT Image 2.6 — supposedly the strongest at Arabic writing — added stray dots to «الخيال». The author was unimpressed overall and still hand-letters text in real design work.
Related event: Arabic image model showdown ends with GPT stumbling on Arabic text(2 posts)→
More from Multimodal
- AI-made music video "Slopcore, Encore" enters Claudia contest via Claude skill + Fable 5.2 — AIandDesign · 2026-11-11
- TRIAD benchmark: no multimodal retrieval system leads across its six subsets — _reachsumit · 2026-11-11
- Vision tokens far denser than text: paper traces VLM modality gap to early-layer processing — _reachsumit · 2026-11-11
- Fable 5.2 one-prompt short film "936": "make a video that will deeply change my life" — yuwen_lu_ · 2026-11-11
- Fable 5.2 renders a photorealistic snowy Manhattan night, brush strokes all AI-made — yuwen_lu_ · 2026-11-11
- Fable 5.2 renders voxel-style axe chopping a log, sparking '3D is solved' claims — adonis_singh · 2026-11-11