Beginner Basics By Beginner 3 - AI Tool
This article is for the Beginner and assumes previous parts have already been read.
The name "AI Tool" can mean more than one thing in the world of AI, for the AI industry at large it can generically mean a software that provides a solution. In other words, it is a software tool that is based on AI. This is obviously NOT what we are referring to in the context of Tensor Art (TA). Here, we are taking about a tool we are using to do image generation, which is what this article will be about. Therefore, If you are talking to someone else outside of TA, do note that "AI Tool" can mean something different from what we are talking about here.
An AI Tool in TA is typically something a user created using ControlNet. ControlNet as previously stated is not for Beginners, so without delving into ControlNet itself or its uses, for now you can think of ControlNet as a programming tool to create AI Tools. Tools generally are meant to enable or make things easier, so the same goes for AI Tools. For the Beginner, AI Tools offer a simplified way to achieve image or video generation. Therefore, before you try or at least try in earnest on your generation page, AI Tools may be a good starting point to let you know what you can do with a model - motivates you to try the model on your own. Remix is another but I will cover that in another part.
Some AI Tools simplify prompting to such an extent as to give you multiple choice options to click on to serve as prompts. Once again RealAnime has this feature. If you do not want to fill in the boxes, there are a few clickable options for subject and another few options for action.
There are many AI Tools that simplify the options and even break the prompting into entry boxes for you to fill in accordingly, kind of like filling in a simple form. Fill it, click go and that is it. For example, currently there is a RealAnime AI Tool that has this feature where it ask you to fill in one box who you want to appear as the subject, and in another box fill in what you want the subject to be doing.
Other AI Tools may allow you to upload images to generate images with or without the form filling way of doing prompts. An example of this is the 3DVideo where you upload a image and it makes it into a rotating view point video using your subject as the focal point. The effect is to make your image into a 3D version with a camera going around your subject. In 3DVideo, its does not go fully around, just rotates little towards the left. This example includes an inpaint option, The inpaint if you were to click on it allows you to use your cursor to "paint" your image to identify for the AI Tool what is the subject that is the focus. The paint tool includes an undo last paint and a reset all paint to zero. You can choose thicker or finer brush sizes, although the sliding scale control seems to be hard to control (for me it jumps a little when dragging). The inpaint tool also tends to have a small viewer making painting fine areas harder. I use the browser zoom to enlarge but beyond a point the left right control to pan the image seems to fail so I have to zoom out to pan then zoom in again to do fine painting. The need for fine painting seem to vary by AI Tool and by your source image, decide for yourself if you want to spend more time fine painting or just painting broadly. Sometimes broad painting is good enough - in other words its fine to "paint outside the lines" so to speak. In the case of 3DVideo, you can even choose not to paint at all and let the AI Tool try to detect the subject. Painting just gives you more certainty its correctly detected.
AI Tools may have some unstated limits. In the 3Dvideo example, a limitation is that it handles landscape (horizontal) images better. Square ones seem to work, although there may be more distortion compare to landscape. Portrait (vertical) images may have cropping and more distortions.
Furthermore, AI Tools often simplify other aspects of generation for the Beginner as well. For example, choosing LORA(s), negative prompts etc. You could want it for simplicity, or it may be at a time where its difficult to do on your own. For example, when Flux (a model) first came out, there was a lack of LORAs. Furthermore, its use and behavior was a little different from other models that have a been around for a while. Therefore, in those early days of generation with Flux, AI Tools usage was the go to method for many users.
AI Tools are written by different people who are targeting different users, have different ideas as to what is desirable in an AI Tool etc. Hence, simplicity vs flexibility, type & quality of generation etc all vary.
One thing to note, is that using an AI Tool is no guarantee that the generation will be a success.
- See what others have used it for to know what it is suitable for.
- Check the quality of results by others to know what to expect.
AI Tools cost credits, each tool has its own cost, so look before you leap. Sometimes there is a discount going on and you can use it cheaper - which is also means the price will go up later, hence be aware and not end up surprised by having paid after the fact.
AI Tools also have a queue just like your generation page. I generally advise only to do 1 generation at a time and not queue generations for an AI tool, as it may fail. Yes, AI Tools can fail. Read error message. If its due to the processing and not due to something indeterminate or what you input, then you can try again if you wish - although there is still no guarantee of success. Having had only 1 generation in queue, and if you did not refresh or close browser etc, whatever you set up for generation should still be there in the queue (try tab), so just click to generate.
Generations from an AI Tool will appear at your generation page so there is no need to track it at the AI Tool page. If you post your generated result, it may appear at the AI Tool as a post too.
