
It puts a workflow community, one-click...
Use one-click generation when you want fewer nodes, or open online ComfyUI when you want to build the flow. Both can polish a prompt in one click. The workflow community lets you search and run flows other people published. LoRA training means uploading a set of images, tagging them, and filling in a trigger word. A private model becomes public if you publish it with a workflow. Compute points need a top-up when they run out, and the daily login grant clears at midnight. The platform is still in a closed test, and you are sent to login if you are signed out. It fits image authors who want workflows and their own models in the browser.
Qingcheng Aidraw is an online image platform from Beijing Qingcheng Jizhi Technology, and it is still in a closed test. It puts a workflow community, one-click generation, and online ComfyUI in one place, so you can run workflows and train your own LoRA without a local install. The page offers text-to-image, image-to-image, inpainting, and upscaling, and Chinese prompts can be polished into English in one click. Generation uses the in-house Chitu engine. You must log in. Compute points come from a daily login grant and from top-ups, and the daily grant resets the next day.
Difficulty: Advanced
An online workflow community
Search workflows other people published and run them without installing ComfyUI on your own computer.
One-click generation and polish
Both one-click generation and the ComfyUI page can polish a Chinese prompt into English before you generate.
Train a model in the browser
Upload a set of images, finish tagging, add a trigger word, and train a LoRA on the site.
Inpaint and upscale
Besides text-to-image and image-to-image, the page offers inpainting, upscaling, and super-resolution.
Running a community workflow
A creator finds a workflow in the community and runs it after login instead of building every node.
Training a model on your own images
An author uploads a consistent set of images, tags them, sets a trigger word, and trains a reusable LoRA.
Polish Chinese, then generate
Someone who writes in Chinese polishes the prompt into English, then generates in one-click mode or ComfyUI.
Pricing may change. Please refer to the official website for the latest pricing details.
No. If you are signed out or the session expired, the page sends you to login. A daily login grants compute points, and that grant clears at midnight. If the balance is too low, top up before generating or training. The site says it is still in a closed test.
Prepare images. The training page asks for images or a dataset in jpg, jpeg, or png. Tag files can be txt or jsonl and must match the images one by one. The trigger word can contain letters, digits, and a few symbols. Check that you have enough compute points and model-library space before training.
Yes. The page says a private model included in a workflow becomes public when you publish, and an already published private model is swapped for the public one. Published models no longer use your library space. Leave a model out of the workflow if you want it to stay private.
Real reviews and feedback from users