Why not just use the strongest general model?

Because text is a separate skill. Nano Banana Pro follows staging instructions better; Qwen renders characters correctly. For a frame with a Chinese sign in it, the second matters more.

Does it work for English text too?

Yes, and well. The gap is widest on CJK, but Qwen is a solid choice for any text-heavy frame.

Is it available inside a reel?

Not currently — the reel pipeline publishes GPT Image 2, Nano Banana Pro and Seedream 5.0 Pro. Qwen is available for standalone image generation, and its output can be uploaded as a reference.

Qwen Image 2.0

Qwen Image 2.0 is Alibaba's image model, and the strongest option in this catalog for rendering text inside an image — particularly Chinese, Japanese and Korean, where most Western-trained models produce characters that look right and mean nothing. Title cards, signage, captions and posters are where it earns its place.

Start creating View showcase

Models

Specifications

Model Specs

Max Resolution

Speed

~7s per image

Price / Credit

24 credits / image

Capabilities

Supported Features

AI Text to Image Generator

AI Image to Image

How It Works

Key Features

CJK Text That Reads

Most image models treat Chinese characters as shapes and produce plausible nonsense. Qwen was trained where that is not acceptable, and it shows on any frame that has to carry a word.