Artisse's new app generates cool photos with the user's face
Miscellaneous / / September 07, 2023
A couple of clicks, and the neural network will “embed” you in any frames of your choice.
A new Artisse app has appeared in the App Store and Google Play. It uses its own generative model to create photos with the user's face from a text description or a photo example.
Like other applications of this kind, before starting work, the user needs to train the neural network: feed it 15-30 of their photos so that it analyzes the features and proportions of the face. It is recommended to take photos from different angles, where there are no strangers or animals, and the face is not covered by a smartphone or something else.
Screenshots of Lifehacker
After sending photos, you can forget about the application for 30-40 minutes: the neural network does not learn quickly. If you allow notifications to be sent, the application will write itself when everything is ready for use.
As already mentioned, the application works according to two schemes: based on a text description or reference photos. In the first case, everything is more or less similar to any neural network for creating pictures: you drive in a request with all the necessary parameters, select the number of frames you want to get at the output (from one to four), and press "Create". As in the case of training, the generation will take 30-40 minutes.
Screenshots of Lifehacker
The resulting photo can be edited, regenerated or saved to the device. Artisse also offers to immediately send the image to the social network - including your own, inside the application.
The "Create from Image" mode basically just inserts the user's face into the selected photo. Technically, the application allows you to adjust the degree of change in the picture, but in practice it has little effect. For example, take this shot from Neon Demon:
And here's what we get with the settings "Slight variation" (left) and "Strong variation" (right):
There is minimal freedom in the details, but the result in both cases is very close to the original. If someone saw a frame or photo that you took as a basis, accusations of plagiarism or photoshopping your face on someone else's picture cannot be avoided.
The application is shareware. At the start, the user receives 25 coins - each allows you to generate one photo. This is enough to process 6-25 requests, depending on the selected number of images. For further work, coins will have to be bought. 10 pieces cost 199 rubles, and the largest package for 565 coins costs 9,990 rubles. You definitely cannot do without a purchase if you decide to retrain the neural network: it costs 30 coins.
So far, the app is only available for iOS and Android, but the developers plan to launch the web version in October.
MUMU LABS
Price: Free
Download
Price: Free
Artisse
Price: Free
Download
Price: Free
More generative neural networks🦾
- The BaiRBIE me neural network will generate a Barbie doll based on your photos
- Stable Doodle neural network launched. She creates pictures from rough sketches
- 10 best neural networks for drawing online