Qwen Image 2.1 AI Image Generator
Qwen Image 2.1 online: one fast model for text to image and photo edits with up to 8 references, free to try.
Sample images
A cozy café chalkboard menu, hand-lettered in white chalk. Top line in large letters: 'VIDLARK CAFÉ'. Below, three neat lines: 'Today's special: Lime Matcha Latte $5', 'Fresh croissants · 7–11 am', 'Free Wi-Fi, good vibes'. A small fuzzy lime-green felt bird figurine with a round black eye sits on the wooden frame of the board. Warm morning light, shallow depth of field, realistic photo.
Qwen Image 2.1 is the compact member of Alibaba's Qwen Image family, released in September 2026: a single model that both creates images from text and edits photos, built on a 7B visual generator so results come back quickly. On Vidlark it takes up to 8 reference images, renders at 1K, 2K, and costs 30 credits at 1K and 35 at 2K. Free accounts can use it with the sign-up credits.
Updated
At a glance
- Made by
- Alibaba
- Price
- From 30 credits per image
- Size
- 1K, 2K
- Input
- Text or your photo
- Aspect ratio
- 1:1, 4:3, 3:4, 3:2, 2:3, 16:9, 9:16, 21:9, 9:21
- References
- Up to 8 images
- Per run
- 1–4 images
- Available on
- Every account, free included
- Account
- Free account required, 100 credits on sign-up, no card
- Unlimited plan
- Free at 1K
Vidlark is an independent platform and isn't affiliated with Alibaba.
Qwen Image 2.1 vs Qwen Image 3.0
Qwen Image 2.1 is 30 credits at 1K and 35 at 2K. Qwen Image 3.0 (30 at either size) is the stronger pick for long text and mixed Chinese and English, and 3.0 Pro (40 at 1K, 55 at 2K) adds finer detail.
| Model | Best for | Top quality | Per run | Edit photos | From |
|---|---|---|---|---|---|
| Qwen Image 2.1This page | Fast text to image and multi-photo edits in one model. | 2K | 1–4 | Yes | 30 credits |
| Qwen Image 3.0 | Clean Chinese and English text in images. | 2K | 1–4 | Yes | 30 credits |
| Qwen Image 3.0 Pro | Qwen Image with finer detail, up to 2K. | 2K | 1–4 | Yes | 40 credits |
How to use Qwen Image 2.1
- 1
Describe the image
Type a prompt, or upload a photo to edit.
- 2
Pick the settings
1K / 2K; the price shows before you start.
- 3
Generate and download
Download the images when they are ready.
What it does well
One model for making and editing
Text to image and photo edits run on the same model, so an image you just made can go straight back in as a reference for the next change.
Up to 8 reference photos
Bring a product, a person and a place together from separate photos, or keep one character the same across a whole set.
Clean English text in the picture
Signs, menus and packaging labels come out with the words spelled as written, down to small print like weights and brewing times.
What is Qwen Image 2.1 Turbo?
Qwen Image 2.1 Turbo is a faster version Qwen released on October 9, 2026. It uses the same 7B visual architecture as 2.1 and generates or edits an image in just 8 denoising steps. Its open weights are on Hugging Face, with ComfyUI files from Comfy-Org, and its API runs on Alibaba Cloud Model Studio. Turbo is not on Vidlark yet; this page runs the standard Qwen Image 2.1. Running the weights yourself needs a CUDA GPU with PyTorch; Qwen publishes no minimum VRAM, and the weights come under the Qwen Research License.
Settings and prices
Each image has a fixed price; the button always shows the exact total before you generate.
| Setting | Price | Example | Unlimited plan |
|---|---|---|---|
| 1K | 30 credits / image | 4 images = 120 credits | Free |
| 2K | 35 credits / image | 4 images = 140 credits | Uses credits |
Prompt tips for better images
Say what must stay the same
When editing, name what to keep ("same face, same jacket") as well as what to change; edits are more precise when both are written down.
Number your references
With several photos, refer to them in order: "the bottle from image 1 on the table from image 2, in the light of image 3".
Draft at 1K, finish at 2K
Settle the layout at 30 credits, then switch the quality to 2K and run the same prompt once for the sharp version.
Example prompt
A ceramic coffee cup on a linen tablecloth by a window, soft morning light, shallow depth of field, 85mm product photo.
Use this promptWhat people make with it
01
Products on new backgrounds
Upload a packshot and describe the scene: a kitchen counter, a beach towel, a studio sweep. The product keeps its shape and label while everything around it changes.
02
Character sets
Feed two or three photos of the same mascot or person and ask for new poses and places, for comics, stickers or a social series.
03
Fast drafts before a final model
Try ten compositions at 1K, keep the one that works, then rerun its prompt at 2K or on another model for the final.
How it compares
How it lines up against similar image models on this site; every one uses the same credit balance.
| Model | Best for | Top quality | Per run | Edit photos | From |
|---|---|---|---|---|---|
| Qwen Image 2.1 | Fast text to image and multi-photo edits in one model. | 2K | 1–4 | Yes | 30 credits |
| Seedream 5 Flash | Cheap, quick Seedream drafts at 1K or 2K. | 2K | 1–4 | Yes | 30 credits |
| FLUX.3 Klein | Fast, low-cost images and photo edits from a new open model. | 2K | 1–4 | Yes | 30 credits |
| Muse Image | Cheap square images with readable text, from a prompt. | 1K | 1–4 | No | 30 credits |
Other models to try
All models
Qwen Image 3.0UnlimitedStrong at Chinese and English text inside images.from 30 creditsQwen Image 3.0 ProHigher-detail Qwen Image up to 2K.from 40 credits· Paid plans
GPT Image 2.5 FlareUnlimitedOpenAI's GPT Image 2.5, balanced and fast: accurate text and precise edits.from 40 creditsGPT Image 2.5 SunburstUnlimitedThe high-precision GPT Image 2.5 for detailed commercial images.from 40 credits
Nano Banana ProUp to 8 reference images for consistent characters.from 55 credits· Paid plansNano Banana 2UnlimitedFlash speed with Pro-level consistency.from 30 creditsQwen Image 2.1 questions
How can I use Qwen Image for free?
Sign up and Qwen Image 2.1 runs on the 100 free credits, with no card: a 1K image is 30 credits.
Is Qwen Image 2.1 censored?
On Vidlark every prompt and uploaded photo goes through a content check before it reaches the model, and adult or harmful requests are refused without a charge. Ordinary portraits, products, art and design work run normally.
Qwen Image 2.1 vs Qwen Image 3: which should I pick?
Pick 2.1 for fast drafts, short English text on signs and labels, and edits that mix several photos (up to 8 references here, 30 credits at 1K). Pick Qwen Image 3.0 (30 credits at 1K or 2K) when the picture carries long passages of text or Chinese and English together, and 3.0 Pro for the finest detail.
Does Qwen Image 2.1 make 2K images?
Yes. The model is built for native 2K output; here 1K is 30 credits and 2K is 35 credits per image.
What sizes can it make?
1K, 2K, in square, portrait or landscape. Each run makes 1–4 images.
Credits, refunds, watermarks and commercial use work the same on every model. See the pricing FAQ
Make a Qwen Image 2.1 picture
100 free credits when you sign up. Failed runs are refunded.