# #
» Blog » How to manage the content of images created by a neural network

img

In this post we will talk about subtraction cues, the so-called « negative prompts », try to outline how to direct the neuronku to the cherished idea and influence the final product image.


24AI allows you to create multiple product images with different backgrounds from a single product photo. The result is impressive: the renderings produced by the neural network can be safely used in product cards and in marketing. They sell, involve, create additional product value. And all this is easy, without special skills and professional software, in a browser, in a few clicks. Previously only designers could do this, but now anyone can do it with ease.

On the agenda:

  • How to « put » objects if they are « lying around »
  • how to use negative prompts
  • tips for making prompts
  • upscaling

Standing or laying, that’s the question

If you specify the phrase « standing on » or « stands on » in the description, the neuronka is more likely to produce images in which your product is standing on the surface with a deep space background, rather than lying down and looking as if it is taken from above.

How to manage the content of images created by a neural network
« Standing on »

If you ignore this specification and leave it to chance, the product shot in a strictly frontal perspective will more often put the neural network on a « flat » environment without depth.

How to manage the content of images created by a neural network
Default positioning

Specifying this combination « stands on, » however, does not guarantee the vertical position of the item in the frame. In general, any refinement in the query, as well as the above, only increases the frequency of output of what you need.

Take pictures of the volume
If you don’t have a picture of the product yet, and you just want to take one, pay attention to the angle in which you shoot it. Show it not strictly from the front, but also from above, and the neural network will have no reason to hesitate to place the product on the renderer.

How to manage the content of images created by a neural network
When you shoot product, show it not strictly from the front, but also a bit from above

If you’re having trouble getting the results you want, email us at hello@24ttl.net and send us pictures. Feedback is always welcome and we will definitely help you.

Negative prompts

On the paid rate, in addition to pre-prepared environment themes, it is possible to create individual queries, add references and deductive cues aka « negative Prompts » or « negative queries ».

It is difficult to imagine in advance what should be absent from the image, until the neural network generates that very something in the image. For example, you can try to subtract the plants in the office, so that they are more likely to be absent from the image. This is how we did it:

How to manage the content of images created by a neural network
Negative prompts « Bag standing in office No plants »

Combine positive prompts (environment descriptions) with negative prompts, and you can better control the content of the final images that 24AI creates for you.

Prompts for making up prompts

Try to follow the recommended query structure. Separate each semantic block of the query from the others with commas.

  • The position of the product (where or on which it lies or stands)
  • Additional elements surrounding it, and their properties
  • Style, weather, mood or time of day
How to manage the content of images created by a neural network

Importantly, the less the neural network applies the parts of the query you specify in the final image, the less familiar it is with what you describe, and the more detail you give it.

24AI is constantly learning, but at any given time it is limited to the amount of visual data it has explored. If you consider that the level of « insight » of the neural network directly affects the images it generates, it will be easier for you to formulate a target query to which it will respond adequately to your expectations.

Experiment. This is the only way to be sure of the strength of 24AI, to bypass its weaknesses, to invent tricks, and to understand the principle of artificial intelligence thinking in order to eventually obtain the coveted images.

Upscaling

All you need to know: upscaling is an increase in the resolution of an image, namely the number of pixels per square inch, by neural network algorithms. The technology is young, but it can already do something. Take a look at how it works in 24AI:

How to manage the content of images created by a neural network
How to manage the content of images created by a neural network

We sincerely hope you enjoy 24AI’s features. Share your generated images on social media by tagging #24aitech. Tell your friends about the service. We look forward to your feedback and suggestions hello@24ttl.net

We use cookies to get statistics that help us improve the service is for you in order to personalize services and offers. You can read learn more about cookies or change your browser settings. Continuing to use by using the website without changing the settings, you consent to the use of your cookies.