The most dazzling AI photo editing tool of 2024 has arrived!
When it comes to revamping your "Youth Memoir Album," you'll be saying, "Is this vibe authentic or what?"
Don't trust anyone, and don't let anyone trust you.
Don't worry, in the next second, there was a sudden "pop":
...What's going on?
This is the latest craze in the photography world - the AI editing tool called Super Canvas.
It's said to be able to create stunning, creative edits in minutes with just a single prompt, without the need for training specific models with multiple images beforehand. You can simply open it up, upload a photo, and start editing.
Recently, it underwent a rapid iteration and upgrade cycle, making waves in the community. Some users have even hailed it as a game-changer, especially for handling quirky client demands, drastically reducing costs while increasing efficiency.
For clients like the Service Scarf Auntie, the Super Canvas's live drawing effects are more than sufficient:
This remarkable AI tool, Quantum Bit, is definitely not to be missed.
Upon delving into it, I discovered that this AI not only excels at full-image editing but also offers the ability to edit specific parts. And believe it or not, it can even change hair color with just one click!
Even if it's just a casual floral print:
She can effortlessly transform into a desert princess or a Tibetan girl in minutes, and the remarkable thing is, she looks exactly the same in both roles.
In the future, you could just throw on some pajamas at home and easily assemble a perfect grid for your social media feed. How is that even possible? It's all thanks to an all-in-one service dubbed the "Million-Image Retoucher" AI.
After some digging, it turns out the brainpower behind this marvel hails from Baidu Netdisk, offering a comprehensive suite of professional photo editing services.
Gone are the days of elaborate makeup, wardrobe changes, and set design for professional photography. With the Million-Image Retoucher, creating stunning photos is as simple as saying a word.
Here's the gist of how it works:
AI Portraits: The focus here is on preserving the pose and facial features while completely changing clothing and background.
Partial Redraw - Enhancement: This focuses on making localized adjustments to the original image, adding or modifying elements.
Partial Redraw - Background Replacement: This is for scenarios where the clothing is already set, but a new background needs to be created.
Let's take a look at the AI's photo editing abilities.
First up, we have a picture of my friend and Jun Ji-hyun, taken in Zhongguancun. What follows immediately catches our eye: the actions and outlines of the individuals in the photo are effortlessly outlined with just a click.
Below are the condition control functions of the super canvas: action capture and outline control.
In addition to these two, there are also depth of field, style reference, and detail enhancement. All five condition controls can upload control maps and set intensities to control generation, with the default being original image control. Up to three condition controls can be combined for each generated image.
Next, on the right side of the interface, there are two optional base models: the Realistic Photography Model and the Film Photography Model. The former leans towards a lifelike style, focusing on details, while the latter incorporates depth of field effects, emphasizing atmosphere.
The prompt supports both Chinese and English, with adjustable weights for each word. There are also inspiration associations for users to choose directly. In the advanced settings, there is a series of controllable parameters as well as reverse prompts; the generation settings allow modification of canvas size and the number of generated images.
I have to say, compared to most domestic AI image editing/generation tools, SuperCanvas offers quite a lot of details that users can customize.
This whole set of settings can be DIYed by users with professional demands and saved as "My Formula".
For those lacking inspiration, the official also provides ready-made "Inspiration Formulas", and the matching between the formulas and the actual generated scenes is also very high:
We opted for a desert-inspired formula with a hint of camel bell vibes, then hit generate. The original Zhongguancun Fenxian photo underwent a swift transformation.
Voila, even mom exclaimed it's real. Consistency in facial features and body posture: accomplished.
Let's take a look at the section for partial redraws. Click on "Edit Selection" to see two functions: one for making changes to specific elements and another for easily changing the background.
Modify local elements where necessary, adding, deleting, or altering elements as needed.
The one-click background change feature is suitable for those who already have a complete outfit and only need to create a scene.
For more detailed gameplay, the official has provided a tutorial on Zhihu. You can click on "Help Documents" to view it.
The document not only elaborates on the effects of each controllable component but also provides some handy tips:
The Aesthetic-Savvy Mega Model
Technically, behind the Super Canvas lies a team from Baidu's Research and Development Department, leveraging the Baidu Intelligent Cloud's Thousand Sails Mega Model Platform to construct an aesthetic AI image generation mega model.
With the support of this mega model, the Super Canvas boasts some remarkable features.
Firstly, the model excels at generating high-quality portrait photographs. It allows for fine-grained control over aspects such as subject posture, image structure, depth of field, lighting, facial makeup, by utilizing a reference image as a control condition.
Maintaining the similarity between the final piece and the original image, the SuperCanvas boasts a killer feature: an intelligent facial fusion algorithm.
This algorithm incorporates user facial information as control conditions in the image generation process.
By analyzing features like facial lighting, skin tone, and texture in the generated canvas results, coupled with the user's input image containing facial features and shape, it seamlessly preserves the user's facial identity in the generated images.
Comparison between Fusion Images and Source Materials
In addition to authenticity, the Baidu Cloud team has also imbued its understanding of "beauty" into the creation of the Super Canvas—aesthetic evaluation model.
To align the model with human aesthetic cognition, the team has constructed a model optimization pipeline based on user feedback, thereby making the generated images more in line with users' aesthetic preferences.
It can be said that through these three models, the Super Canvas provides a one-stop portrait photography retouching solution for photographers.
Of course, behind the "superpowers" of these models, our familiar old friend—the Baidu Intelligent Cloud Qianfan Large Model Platform—provides indispensable support with its powerful underlying models and computing power.
The Qianfan Large Model Platform is a one-stop large model development and service operation platform for enterprise developers, providing various large model services, as well as a variety of AI development tools and a complete development environment, allowing developers to complete application development in one place.
"Photography ChatGPT" is on the horizon?
Returning to the Super Canvas, the new experience it brings has led users who have tried it to exclaim: the era of ChatGPT for photography is coming.
This also raises the question: What is the "ChatGPT moment" of photography?
A year ago, domestic large-scale model competitions ignited the language model arena. Today, image generation seems to have become a battleground for domestic large-scale model manufacturers.
The emergence of Super Canvas can be seen as a fresh perspective entering this battle. While AI retouching and AI image generation are common, the area of "GenAI retouching" still seems largely untapped in China.
To be more specific, Super Canvas targets the pain points of domestic photographers.
Industries like photography, which are not closely tied to AI, may not have access to tools like SD and MJ. Super Canvas, being a domestic application, carries the weight of being "made in China," a factor that resonates with many users and makes it more accessible.
When photographers encounter clients who deviate from the norm, Super Canvas can help them save a significant amount of retouching work with just one click, easily meeting the needs of cosplay portraits, client photos, commercial videos, and more.
It is understood that as the product continues to be optimized and upgraded, Super Canvas can also handle needs such as secondary sales of client photos and repeat business from old clients.
For example, offering AI-generated portraits as a gift with a portrait session is indeed a clever way to attract customers, isn't it?
The super canvas is more convenient in that it requires no training, deployment, or downloading and installation. This eliminates the technical barriers of setting up workflows, as well as the hardware hurdles of deploying models.
All you need to do is open the webpage, and you can start using it right away. With its simple and intuitive interface, it's accessible to both AI novices and experts alike. Any photographer, regardless of skill level, has the opportunity to use it and tackle various modern scenarios without feeling helpless.
If you're an advanced AI user, there's a chance to explore more diverse and high-level gameplay in the app by freely combining control conditions with your advanced prompting skills. Feeling intrigued? Why not give it a try yourself?
