MiaoHua

MiaoHua

Miaohua,SenseTime,7-billion-parameter Artist painting large model developed by SenseTime

Visit Site →

Official Website of MiaoHua, Sensetime, powered by the self-developed Artist painting model with 7 billion parameters

About MiaoHua

MiaoHua is a free AI painting creation platform launched by Sensetime based on the self-developed Artist painting model with 7 billion parameters, making it easy for users to customize and generate images effortlessly. MiaoHua supports text-to-image and image-to-image prompts, combining precise control and style models to generate high-quality artwork content anytime, anywhere.

MiaoHua Official Website:https://miaohua.sensetime.com/zh-CN

Main Features

Product Features of Sensetime MiaoHua
Text-to-Image. MiaoHua can generate corresponding images with just a prompt, with the highest resolution up to 6K UHD.
Image-to-Image. Users can provide images to transform and generate similar images in other styles.
ControlNet Precise Control. Supports pose control, line coloring, depth detection, and more.
Custom Training LoRA Models. MiaoHua also supports users to upload local images, combine with Sensetime's self-developed painting model or open-source models to train customized LoRA models for efficient generation of personalized content.

"MiaoHua SenseMirage," the image generation platform under the SenseNova large model system of Sensetime, has officiallylaunched the mobile version, and is now free to use.

More dynamic interface, no compromise on features.

MiaoHua mobile is designed for mobile usage habits, allowing users to start text-to-image creation anytime, anywhere by logging into the Sensetime MiaoHua official website via mobile.

At the same time, MiaoHua has also updated the PC interface: clearer function layout, more intuitive navigation, and simpler operations. By clicking "Start Creating" on the homepage, users can quickly start drawing, with maximum user-friendliness.

With multiple upgrades, Sensetime MiaoHua has officially entered version 3.5, understanding aesthetics, writing words, and customization, three skills to open up infinite creative space.Not an aesthetics genius? No time to train aesthetics?

Don't worry, the 3.5 version of Sensetime MiaoHua

has built-in a new aesthetic guidance system, helping you master this "mysticism" of aesthetics.For three-dimensional paintings, MiaoHua can balance artistry and delicacy. While showcasing professional photography-level depth of field effects

at the same time, it retains the painting's sense of hierarchy and texture details. Especially forAsian portraits, MiaoHua has also optimized, reducing the feeling of skin smoothing and detail loss, and accurately depicting the beauty of the East.Fortwo-dimensional art

, MiaoHua also continues to optimize: while realizing various stylesat the same time, it can achieveultra-high color harmonyin both warm and cold tones. The brush strokes are delicate and smooth, and the detailed depiction can match the level of professional illustration.MiaoHua is responsible for adding the finishing touch, and you are responsible for "WOW!"If AI painting is "alchemy", then ControlNet is the shortcut to "alchemy".With just one reference image, ControlNet can help players extract the outline lines, depth of field information, composition information, and image content information of the scene.

Based on these information, carry out

other style re-creations

,

efficiency can be greatly improved.The new version of MiaoHua has integrated this treasure tool, without the need to jump, and can be used directly under the "Image-to-Image" function.It brings more creative experiences to advanced players.Understanding aesthetics

,Prompt writing is no longer difficult. Through MiaoHua, users only need to input a few prompts, and it can automatically complete them. At the same time, the built-in aesthetic guidance system optimizes the visual level synchronously.

Customizable

,Simple and fast to achieve personalized expression. MiaoHua's convenient LoRA model training capability allows users to customize personalized exclusive models in just 10 minutes by dragging and dropping local images, using Sensetime's self-developed painting model or open-source models. Users can also import third-party models with one click, avoiding local deployment, enjoying Sensetime MiaoHua's inference acceleration service, and greatly improving generation efficiency.Inspiration Pool

,Help you explore creativity. For professionals in creative vertical fields such as advertising marketing, film and television special effects, game development, design, photography, and more, MiaoHua can provide high-quality image materials and sources of inspiration, helping users efficiently build prototypes of works and comprehensively improve creative ability.In addition, Sensetime will also open the API interface to B-end customers, providing convenient and efficient services for enterprises and creating diverse AI painting communities for different user circles.

Currently, MiaoHua has servedmore than 60 enterprise-level customers, covering industries such as construction, marketing, education, entertainment, games, automotive, cultural and creative industries, with nearly one million users registered and experiencing the web-based product, with over 100,000 daily calls.In the future, "Sensetime MiaoHua SenseMirage" will continue to increase investment in core technology, community construction, and platform functions,connecting "generate + optimize + share" one platform full chain, continuously providing users with better experiences, allowing artistic creativity to flow freely, and releasing more potential and value of AIGC.

 


 

In two seconds, you can turn your imagination into "reality".

At the Sensetime Technology Day, Sensetime's "SenseNova" large model system was officially launched, and the AI content creation community platform"Sensetime MiaoHua SenseMirage"also made its debut.Sensetime Chairman and CEO Xu Li demonstrated the painting capabilities of "Sensetime MiaoHua SenseMirage" based on Sensetime's self-developed AIGC model on-site; also introduced the MiaoHua platform based on Sensetime's large-scale GPU computing power support, which can help usersavoid the process of local deployment

, both using Sensetime's self-developed painting model to efficiently generate high-quality content and also importing local models and other third-party open-source models with one click to generate more diverse content.

MiaoHua also provides

specific reasoning optimization services, and after importing open-source models, they will automatically adoptMiaoHua's underlying model compilation technology for acceleration
, and the actual test shows that the image that needs 10 seconds to generate on a local RTX3070 graphics card can be generated in just 2 seconds through MiaoHua's acceleration technology.In addition, MiaoHua also supports users to upload local images, combine with Sensetime's self-developed painting model or open-source models to train customized LoRA models for efficient generation of personalized content.(LoRA, used for further training with small data, adapting the capabilities of the large model to a specific vertical field, but affecting the generalization ability of the large model)
(Click the image to view it in full size)· MHTeam: MiaoHua Team, the creative model of the Sensetime MiaoHua team· Latest Creations: The latest paintings created by users using the latest Sensetime self-developed painting model and shared on the square· Selected Models: Users can add Sensetime's self-developed models or open-source models to their "Selected Models" list by collecting, which is convenient for access and calling when needed
· Civitai: A community open-source model sharing website
· Hugging Face: An AI open-source software company, mainly providing model hosting and GPU computing power support services

The "Sensetime MiaoHua SenseMirage" platform brings convenience to all kinds of creators and can also fully leverage community power to gather content creativity, promote the construction of the ecosystem of AI text-to-image technology and application through content co-creation

. .

.

.

.

.

...