Qwen Image 3 Text to Image: lock this still before you buy motion (2K, 4cr)
Qwen Image 3 Text to Image is the stills row. Animate after approval. Do not generate the pack inside the video model.
Qwen Image 3 Text to Image is the stills row. Animate after approval. Do not generate the pack inside the video model.
The catalog is specific: Qwen's third-generation image model. Strong prompt adherence and complex text rendering, in both Chinese and English, with intelligent prompt rewriting on by default. 4 credits. 2K max output. Qualities 1K and 2K. Text-to-image. No audio. No duration. Aspects 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16, 21:9.
If the brief has words on the pack, this is the lock step. A video model will invent motion and invent lettering. You will not be able to screenshot the can.
Lock the lettering at 2K
Qwen Image 3 Text to Image is the card you open when the still has to read. Write the exact string you need on the pack, the poster, the storefront. English or Chinese. Prompt rewriting is on by default — your short brief may come back expanded. If a legal line must stay verbatim, say so in the prompt and check the still at 100%. Rewriting is a help for scene language, not a licence to paraphrase a claim.
Prompt adherence is the reason to pay 4 credits instead of 1. Prompt expansion is the default behavior to watch. The text-to-image tool is the door. Qwen's roster also holds Qwen Z Image at 2 credits — a different card. This page is Image 3, 2K, 4 credits.
Set 1K while the words are still in draft. Set 2K when the still is the still. There is no 4K chip on this row.
Chinese and English on the still, not in the clip
Video models are bad typography engines. They will give you seven frames of almost-Hanzi and a bottle that legal cannot ship. Do the bilingual lock here. Approve the characters. Then animate a file whose type already passed.
Aspect belongs on this card. 21:9 and 9:16 are both listed. If the I2V row inherits crop from the image, Qwen's frame is the video frame. Do not lock 1:1 "for now" and crop later through a headline.
Four credits is not a video bargain. It is the price of a readable still. Generating the same pack "inside Hailuo" or "inside Kling" to skip this row is how you buy 6s of unreadable type at a video rate.
Motion waits for the yes
When the still reads:
- Animate on an image-to-video row with this file as the start frame.
- If one object is wrong and the type is right, that is an edit-image job, not a new 4-credit lottery if you can avoid it.
- If the type is wrong, reroll Qwen Image 3. Do not hope motion will hide a misspelled brand.
The best AI image generator list is the shelf. Image 3 is the text-rendering, 2K, 4-credit member — not a motion model with a stills nickname.
FAQ
Will Qwen Image 3 animate the poster if I ask for a camera move?
No. Content type is image. Camera language in the prompt may change composition of the still. Time is a later row.
Why 4 credits when other stills rows are 1?
Because this card's listed job is prompt adherence and complex text in Chinese and English, at 2K. The 1-credit rows are not this typography contract. Pick the row that matches the still.
Can I turn prompt rewriting off?
The catalog says intelligent prompt rewriting is on by default. Check the UI for a toggle on your run. If a legal string must survive, inspect the output still before you buy motion — rewriting is not a reason to skip review.
Should I lock 2K every time?
Qualities are 1K and 2K. Use 1K to argue about the words. Use 2K when the still is the plate you will animate or ship.