Image Battle #1: Five Models, One Antique Pocket Watch
Nano Banana 2, Nano Banana Pro, FLUX.2 Pro, Seedream 5.0 Lite & Pro — same macro prompt, one fal.ai run each. Rankings, per-model notes, real per-execute pricing, and all five stills at the end.
Our first image-model battle is live — a separate track from video models. We sent the same antique pocket watch macro prompt to five text-to-image endpoints on fal.ai: Nano Banana 2, Nano Banana Pro, FLUX.2 Pro, Seedream 5.0 Lite, and Seedream 5.0 Pro.
One attempt per model. No rerolls, no playground cherry-picking. On first pass we did not see hard failures or obvious glitches — the fight is about composition, where your eye goes, dust in the beam, and cost at scale.
Seedream 5.0 Pro took first place on value ($0.0675/run on fal.ai). Nano Banana Pro was the best pure image but at ~$0.15/run — hard to justify for volume. Seedream 5.0 Lite landed last: watch too small, dust felt wrong.
The full write-up — prompt, methodology, ranking, honest notes on each model, and all five images — is on the image model comparison page. Results are at the bottom so you read the reasoning first.
Rankings, per-model notes, pricing, and all five images are on the comparison page
Read full image battle comparisonSee the raw outputs from this test
Open battle archive