How to Make Your First Poster with Xingliu AI If You Don't Know Photoshop

A friend recommended Xingliu AI Design Agent to me, so I tried it to see if it's actually any good. My friend does self-media and said you don't need to know Photoshop, just type out your requirements. I registered an account these past two days, went through it from scratch, and noted down the steps for people who've never touched design tools.

First, let me clarify what it is. Xingliu AI is a design Agent made by LiblibAI (Qidian Xingyu), the domestic version of the overseas Lovart. The word "Agent" sounds mystical, but it's really just an assistant that understands human language, breaks down tasks itself, and asks if you want changes after finishing. Unlike the old tools where you input a prompt and get one image, it can generate images, edit them, and make videos in a continuous flow—it's a complete process.

1. Preparation. You can access it via web or mobile, log in with phone number or WeChat, no invite code needed. New accounts get free credits. My advice is first use the free credits to run through the entire process once, then decide whether to pay. On pricing, public reviews mention 49 yuan for 4000 points, roughly 1 yuan for 80 points—whether it's expensive depends on how much you use it.

2. Getting started. After logging in, click New Conversation. The left side is the chat area, the right is the canvas. The first sentence is crucial—don't just write "make a poster," because then the Agent can only guess. You need to give it a Brief, i.e., a requirements description, clearly stating what to make, who it's for, what style, and which text should appear in the image.

3. Checking results. After sending, it'll first give an approach, then start generating, and the canvas on the right will produce images one by one. At this point, don't rush to export—first check if the composition and text placement are correct.

4. Iteration. If unsatisfied, just speak up, like "make the title text bigger," "change the background to warm tones," "move the person to the left." It edits on the original image, not starting a new one—this saves hassle compared to traditional image generation tools.

5. Export. Click Export on the selected one to get the image file. My first run through the whole process took about ten minutes.

On pitfalls, what I encountered is similar to what reviews say. First, occasionally points are deducted but nothing comes out—in that case, don't keep clicking, refresh the page and check history, most can be recovered; second, estimated time doesn't match actual, it's slower during peak generation times; third, on mobile, the credits and new conversation buttons overlap, easy to mis-tap. The most critical issue is actually writing requirements too vaguely—the Agent can only guess, and the output will be off.

Your need	Don't write it like this	Suggested way to write
Event poster	Make a poster	Vertical event poster, main color dark blue, main title at top, minimalist style
Brand Logo	Design a logo	Make a text-based logo for a coffee shop, combining Chinese and English, mainly black and white
Article image	Add an image	Add a horizontal header image for an article about city night running, realistic photo feel, cool tones

One more thing: it's natively optimized for Chinese, so writing requirements in Chinese is much smoother than translating to English and feeding it—this is also the biggest difference between the domestic and international versions.

The core is just one sentence: treat it as an intern who can do work; if you explain requirements clearly, the quality of its output will be stable. The next step to try is extending the same visual into multiple materials of different sizes, or directly converting generated images into short videos—that's where it really pulls ahead of single-point image generation tools.