Image Generation

Whole subtree ・ 1 votes

Compare text-to-image AI tools, from anime-style illustration to photorealism, rated on their strongest styles and how usable they are for commercial work.

SRecommended

GPT Image (ChatGPT 画像生成)

4.0provisional

AGood

BAverage

CSomewhat flawed

✗Nearly impossible

Rating criteria

※ Your tier is simply your own subjective judgment — that's fine. The following are rough guidelines when voting.

SRecommended
Executes tasks effortlessly.Produces the intended output with few prompts and turns.Excellent cost performance.Almost no hallucinations.
AGood
Experienced users can accomplish most tasks.Eventually produces the intended output.Good cost performance.Few hallucinations.
BAverage
Produces the intended output only after many prompts and turns.Average cost performance.Errors and hallucinations occur.
CSomewhat flawed
Fairly likely to never produce the intended output no matter how many turns you take.Frequent errors and hallucinations.
✗Nearly impossible
Fails to produce the intended output in most cases.

Vote

Compare with the Arena benchmark

arena.ai (formerly LMArena) blind-battle Elo, converted into tiers for reference ・ arena.ai ・ Data: LMArena Leaderboard Dataset (CC BY 4.0)

How to read this tier list (what is Arena?)

Arena (formerly LMArena) is a voting site where people compare two anonymous AI answers side by side and pick the better one. Votes from around the world are aggregated into an Elo rating, the same system used in chess.

Tiers come from the Elo gap to the board leader: within 20 = S, within 35 = A, within 60 = B, below that = C.

In the per-service view, each service is represented by its highest-scoring model.