In the field of design, a revolution triggered by AI is quietly emerging. Recently, an AI design assistant called Lovart has quickly become popular, and with its powerful design capabilities, professional designers have also praised it. It can not only quickly generate design works such as posters, brand VIs, storyboards, etc., but also supports professional-level secondary editing, making design more efficient and convenient.
Another Agent is popular all over the network –
In one sentence, professional visual design is done, even professional designer Big V is crazy Amway!
Painting style be like:
Let’s just say, the whole thing is fashionable.
@Ege the design practitioner who created them said that he used this AI design assistant to easily complete a complete set of fashion brand visual design solutions, while he originally needed to “spend several days” alone.
Some people also used this agent to make Tesla Robotaxi billboards:
It turned out to be liked by Musk himself.
In the official display, there are many cool works, and the following posters are also produced by this Agent:
Even not limited to pictures, videos and 3D creations can also be done:
This agent is also special, supporting a series of professional-level secondary editing, which can achieve image and text separation and layer editing.
How can product managers do a good job in B-end digitalization?
All walks of life have taken advantage of the ride-hail of digital transformation and achieved the rapid development of the industry. Since B-end products are products that provide services for enterprises, how should enterprises ride the digital ride?
View details >
Now you really don’t have to worry about bad text generation.
A mediocre drink picture can be transformed into a beautiful promotional image in minutes:
This agent is called Lovart, and it focuses on poster, brand VI, and Storyboard (script, shot, and storyboard generated by sound and picture packaging).
As soon as the beta version was launched 24 hours ago, the number of people applying for internal testing exceeded 20,000.
Netizens have already started playing, and the results are frequent, let’s feel the popularity of this:
There is a new toy, the qubit naturally can’t be missed, the internal test qualification has been obtained, and the actual measurement has started~
All-round design capability measurement
When you open Lovart, you will be greeted by a seemingly ordinary dialogue interface:
But after entering the command, you will find that it is not simple, and the left side becomes a pieceA canvas with no borders and can be scaled at will。
From initial inspiration to mid-term multi-program iterations, to final task delivery and fine-tuning, the entire design process can be completed on this canvas.
Recently, we are more interested in blind boxes, so let’s first ask it to help design a set of blind boxes and see how it performs.
I saw that it quickly disassembled the prompt to understand the requirements, and then automatically planned to complete the design of 6 images step by step.
Come on, show:
There is also a hidden model, which is more exquisite, and colleagues want one when they see it.
It’s not over yet, and then it “takes it upon itself” to generate a comprehensive layout showing all the characters, and even the box of the blind box is designed:
And looking at the design process of Lovart, Lovart is not a simple drawing, it has its own set of design concepts.
Lovart worked out the design details to the point where each character was designed to characterize, and the entire set was named “Woodland Wizards” before the final generation was carried out.
The series maintains a highly consistent visual style, with all characters carrying iconic elements of the series, and each character is uniquely recognizable.
Let’s take a look at a set of blind box design products, with 3D printing……
Image design This piece of Lovart does have something on it, let’s see if it can really break into the fashion circle.
Ask my girlfriend if she wants to come to the street to take a photo and transform it.
Minutes, “snap”, isn’t this the heroine on the cover of a fashion magazine?
Even if the details are not handled well and the text is wrong, don’t be afraid.
To make a point, compared with general-purpose agents, Lovart specializes in providing various professional-level secondary editing functions for visual design, which can be interacted with through three modes: Talk-Tab-Tune:
- Talk: Assign tasks in natural language directly in the main dialog
- Tab: Click any selection on the artboard to send commands and easily modify them
- Tune: Modification within the layer
Select a picture and click Tab to perform secondary optimization on the image.
Enter in the Talk or Tab dialog box“Generate an editable text version of this poster” triggers the automatic separation of images and texts.
Then you can drag and edit the text on the image by layer.
With the blessing of a series of professional modification functions, AI images finally do not have to become “waste images” because of poor detail processing.
It can handle all types of designs, high-level~
Subsequently, we sent Lovart a qubit logo and tried to ask him to design a brand VI solution.
Somewhat surprisingly, Lovart did not generate it immediately, but designed it on the basis of the original logo, some of which are faithful to the original logo, and some of which have been changed with larger changes.
After selecting a picture, click the “Add to Chat” button below and tell it our choice:
The finished product is out of the oven, and the qubit designer’s job is not guaranteed.
Pick a few of them for the “qubit wild spokesperson” to show, that is, a sentence thing:
In addition, there are a series of tools above the canvas to expand, change backgrounds, remove, add elements, and more.
For example, remove the parts you don’t like on the bag with one click, and apply it wherever you don’t like it:
In addition to this, it can also be used to design storyboards, advertising videos.
We uploaded a mediocre photo of Coke and wanted Lovart to generate a promotional video to drink 25 Coke for 2 pieces.
After receiving the instructions, Lovart began to create, gradually generating a series of storyboards based on this picture, and also generated BGM.
Eventually it automatically assembled it, so we got it with one click:
Okay, okay, that’s how it feels.
However, it is worth mentioning that the generation time is a bit long, with an average of a few minutes to generate a poster, and even tens of minutes to generate a short video. Of course, slow speed is also a common problem in Agent products.
Function revealed, multi-model scheduling
There are more details behind Lovart’s features.
First of all, Lovart’s image, video, and 3D generation capabilities actually come from the fusion scheduling of multiple models.
If you look closely at its generation process, you can find that multiple models work together.
According to the official display, users can actually use GPT image-1, Flux pro, OpenAI-o3, Gemini Imagen 3, Kling AI, Tripo AI, Suno AI and many other top models in Lovart in one stop.
These models are all integrated on such a canvas that users don’t need to switch between platforms and tools back and forth, and tasks are assigned to Lovart, and it will automatically call them for you.
Secondly, Lovart has also built a design process internally, which can automatically disassemble users’ high-level design requirements into a structured task sequence, including style definition, element selection, layout design and final rendering.
This process thinking allows Lovart to think and work like a professional designer, and designers are more native.
whereasAll planning, analysis, and execution in the task process are comprehensively used in the current SOTA-level model.
Among them, the overall planning is revealed to be based on Claude 3.7 data, and I have tuned the new model; The task analysis of nodes is determined by the task characteristics, and closed-source & open-source models such as o3, Gemini, GPT-4o, DeepSeek, and Qwen are comprehensively used.
Behind a series of highly controllable editing functions after generation is the “PS-level” adjustment tool built into Lovart.
These include layers, masks, text fine-tuning, etc., such as layers, which also support fine adjustments such as transparency, hue, saturation, shadows, exposure, contrast, etc.
The final result is delivered, and Lovart can also provide various formats such as PNG, PNG, SVG, etc.
In addition, the Lovart team has responded to the most widely talked about protocols such as MCP and A2A for agents.
They said that they currently use more concise internal protocols to solve specific problems in the design field, and will be compatible with these mainstream ecosystems in the future, improving interoperability with a wider range of AI tools.
Lovart’s brand VI Lovart does
After some in-depth experience, I have to say that Lovart brings a different experience from other AI design tools, which can understand the design process and professional needs, and deliver results that are closer to professional standards.
However, Lovart does not seem to replace designers, but forms a new human-machine collaboration relationship that can effectively handle repetitive work and basic creativity, but high-end customization, strategic design thinking, and core brand decisions still require human designer participation.
So what are its application scenarios in real life?
Perhaps for SMEs without a professional design team, Lovart can significantly reduce design costs, and through simple natural language descriptions, non-design professionals can create professional visual content that matches the brand’s tone.
In a fast-paced marketing environment, Lovart can also generate high-quality new product launches or holiday promotion posters in minutes, helping marketing teams stay ahead of the curve. Especially for e-commerce platforms, which require a large amount of visual materials, Lovart’s batch generation capabilities can significantly improve content creation efficiency.
In addition, about the team behind Lovart, there is a lot of news that Lovart is an overseas product of Liblib, and the qubit has been replied by its team:
Lovart is a new product operated independently by Liblib’s overseas subsidiary, and its main leaders are Wang Haofan and Takumi.
Wang Haofan is a senior algorithm engineer and the core developer of InstantID. The project was also praised by Yann LeCun, one of the Big Three in deep learning.
The Lovart team also directly revealed thatLovart’s brand VI and visual extension are also made with Lovart:
What other scenarios do you think it can be used for? Welcome to leave a message to share. Friends who are interested in Lovart may wish to try it out for themselves, and the official website of LovartAI can apply for internal testing.
Official website link: https://www.lovart.ai/home
