LinkedIn: Tool transforms photo; see how to use – 05/19/2023 – Tech

Design platform Canva’s photo-editing tool lets you take out compromising objects and add appropriate outfits to almost any image. On Twitter, influencer Cara Mia used the tool to hide a bottle of drink and a neckline with a new blouse – the result passes the LinkedIn photo sieve.

The feature is open to the general public but is still in the testing phase. The report used “magic editing”, as the tool is called, and, on the one hand, suffered from bugs such as extra hands, misshapen clothes and mouse selection issues. On the other hand, he got acceptable images, after trying.

The Canva team works to resolve the errors, according to the company’s product marketing manager, Hannah O’Dea. “Because the tool is so easy to use, the way to get the edit you want is to repeat the process,” she says.

Hannah claims that magic editing works better with objects and scenery editing than human portraits.

Image-generating artificial intelligences, in general, have difficulties to deliver proportional hands, feet and heads. Excess or missing fingers may also appear.

Canva’s feature works with Stable Diffusion’s open-source artificial intelligence, which generates images from text. This technology is a competitor to Midjourney, known for having generated the fake images of Pope Francis.

What’s new is that Canva adds the option to use the mouse to create or change elements only in the selected area, according to product marketing manager Hannah.

This feature also comes with a problem: the artificial intelligence can distort selected elements without receiving textual instructions from the user.

The report, for example, selected the entire silhouette of a person in a photo and asked that a suit be added. Canva also swapped the subject’s face with someone else’s.

Hannah says Canva has a strict policy to combat bias. “We encourage and invite users to flag any problems with the generated images, including if they think they are reinforcing prejudices”, says the company in a note.

Another difficulty in using it during the test was the precision of the mouse and possible slowdowns on the platform when selecting the area to be edited.

Hannah recommends that users choose the most suitable brush size to get more detail when editing. According to the manager, more detailed and direct requests deliver better results.

It is better, for example, to ask for a piece of clothing to be created in the selected area, rather than to change the depicted person’s clothing.

The Magic Edition was released on March 23 during the company’s annual Canva Create event. At the time, the platform announced other artificial intelligence features.

Magic Design lets you create images from text, as does Midjourney and Dall-E, from ChatGPT creator OpenAI. Still in the testing phase, the technology can also be generated to create templates for slideshows.

Another innovation is the magic eraser, which erases image elements with AI. This last feature is only available to Canva Pro subscribers. The service costs BRL 289 per year.

HOW TO USE

To use Magic Edit, a person needs to go to the Canva website and create a new design, in the desired dimensions. Next, add a photo. From there, the steps are these:

1. Select photo with one click
2. Select the “edit photo” option in the upper left corner
3. Hover the brush over the desired editing area
4. Describe the object you want to generate
5. Select one of the four images delivered by the platform
6. If none of them are satisfactory, the user can generate four more alternatives with one click