AI Design Tool "Lovart" Major Update Aims to Revolutionize Workflows

According to a report by Lin of iFanr, the emergence of AI image generation models has paradoxically increased the burden on designers. While prompt input and generation itself may be completed quickly, behind the scenes lie hidden manual tasks such as searching for references, downloading, categorizing, explaining intent, and post-generation cropping or layer separation. Lovart aims to solve this problem not by focusing solely on generation, but by optimizing the entire workflow.

Automating Inspiration Collection

The “Inspiration Collection” introduced in this update is implemented as a browser plugin. On major platforms like Pinterest, Xiaohongshu (Red), Taobao, Amazon, and Instagram, simply hovering over an image reveals a “Save to Lovart” button. Saving while simultaneously selecting the appropriate category completes the material organization process. The collected assets, along with design target product information, can be added to Lovart’s chat box, initiating generation that maintains stylistic consistency. Furthermore, Lovart automatically searches for additional online materials during the generation process, allowing users to view results from Pinterest and Google Image Search on-screen.

Enhancing Skills and Autonomous Agents

Lovart operates a marketplace with over 100 “Skills,” provided by both the official team and community creators. These skills cover five categories: marketing materials, brand visuals, social media, product design, and creative styles. Skills package a series of operational methods and delivery standards for tasks like applying effects to logo assets or creating miniature advertisements.

During skill execution, the agent makes autonomous decisions. For instance, with the “Miniature Giant Object Advertisement” skill, it recognizes the provided product details, autonomously designs a miniature world matching the product attributes, and may ask for additional information like image ratio or copy style before constructing the final prompt. Users can also custom-register and reuse their own production processes as skills.

Converting to Diverse Formats

Enhancements extend beyond static image generation to include conversions to videos and HTML files. Generated images can be transformed into videos within the canvas with one click, invoking the Seedance 2.0 model behind the scenes to output in formats like GIF. An image layer separation function is also included, facilitating subsequent design adjustments.

Advanced Brand Understanding

In poster generation tests, providing only a collection of Braun product images allowed the agent to read the brand’s design DNA from the product shapes. Maintaining warm gray backgrounds, hard shadows, and a functionalist composition, it generated a series of images using four different setting methods. For creating key visuals for Huawei smartphones, it first explored similar styles online and presented a design direction board for selection based on four elements: composition, visual representation, color system, and typography. The final image was generated after this selection process.

Expanded Output Formats

Three tier options are available, from the basic “Basic” plan to the upper-tier “Max” plan supporting high-resolution 4K images, 1080P videos, and webpage/PPT output. This simultaneously presents the automation of the entire workflow and the possibility of generating more diverse deliverables.

Editorial Opinion

In the short term, the emergence of such workflow-integrated tools will change the proportion of a designer’s “working” time versus “creative” time. As repetitive tasks like asset collection and format conversion become automated, more resources may become available for client communication and concept development. However, the prerequisite is that AI-generated outputs must maintain the accuracy to precisely capture individual designers’ intent and brand authenticity. In the long run, there is a possibility that the “black-boxing” of the design process could accelerate. The entire flow where agents autonomously select references, decide layouts, and convert formats might make it difficult to later trace the thinking process behind the work. New challenges may arise in skill transmission and building critical evaluation standards within teams. This update can be seen as an attempt to reposition AI from a “design tool” to a “participant in the design process.” However, if a cycle where humans approve every decision made by AI becomes established, it could conversely complicate the decision-making chain and increase on-site burdens in a different form.

References

• “Lovart 悄悄大更新，这一次轮到 AI 适应设计师了”, by Lin — 爱范儿, 2026-09-01T10:28:15.000Z (ARR)
• Source URL: https://www.ifanr.com/1677574?utm_source=rss&utm_medium=rss&utm_campaign=