Transcription
Hello. Recently, the so-called Jon Prompting has been gaining momentum. At first, I thought many already knew about it, however, as it turned out, this new topic requires a separate explanation. So, let's figure out what Jon Prompting is and how to use it for generating text, photos, video, and music. JSON in general is a simple format of structured data that is actively used in programming. But in recent months, it has been increasingly used for creating prompts. The essence of a Jon prompt lies in the clear separation of request elements. It is built on key-value pairs, where the key is the name of the element, and the value is its content. Keys can be set independently, based on the logic of the task, so you don't need to memorize anything from the field of programming. Let's look at an example. Here is a prompt in the usual format, which consists of four important elements, such as role, task, context, and response format. Now you see the prompt on the screen. You can pause the video and copy it. Although such prompts are very understandable to both us and neural networks, but, as practice shows, in JSON format, they give slightly better answers from neural networks, so let's convert it to Jon format. And again, on the screen, you see the updated prompt, which you can copy by pausing the video. And how exactly does the new prompt in JС format differ from the previous one? In essence, nothing substantively, but the difference lies in its structure. Let's break down the new structure. First of all, each line in the new prompt is a key-value pair, where the key is written in quotes on the left, then a colon, and after it, the value is also in quotes. Pairs are separated by commas. The entire block is enclosed in curly braces. This is a mandatory part of the Jon syntax. The order of keys is not critical, but it logically preserves the structure. First, the role, then the task, context, and response format. Such a prompt writing looks more structured, and therefore more unambiguous. The neural network will better understand what you intended and what to expect from it, and you yourself will be able to more clearly imagine the final result. But JSON Prompting offers another advantage. Unlike a prompt in a continuous text format, in JSON format, it is easier to add new elements to make the prompt even more precise. Let's imagine that we want to specify the style and length of the response. To do this, it is enough to add two new elements describing the style and length of the response. Thus, JSON Prompt is easily scalable. You can add any new parameters to it: style, length, format, audience, limitations, and even output language. All of this can become separate keys and their values. So, now that we have understood the basics of Jon prompting, let's look at how this approach works in generating images, video, and music. Suppose you want to generate an image using a neural network like Leonardo or Dali. Instead of a regular text prompt, like the one you see on the screen now, you can format it as JSON, breaking down the description into structural elements. Here's how the new prompt will look. As you can already notice, in such a prompt, you explicitly specify what should be in the image. This is described in the values of the subject and elements keys. When this happens, this information is the value of the key, you clearly define in what style and colors it should be, which is indicated in style and color scheme, as well as a description of the compositional logic in the value of the composition key. With such a prompt, the neural network will no longer have any misunderstanding of what the result of its work should look like, because it does not get lost in the flow of words, it follows the structure and better fulfills your request. This is especially useful in professional visual work, when accuracy and predictability of the final result are important. For example, when the same character should be in different locations, clothes, and settings. Jon Frompt can contain a clear description of characters and locations, and when generating new frames, in most cases, you will only need to change the description of locations and other characteristics, leaving the character description unchanged. Now let's imagine that you are using services capable of generating video, such as S from Open AI or Cleaning AI. Instead of a free description of the scene, you again write a JSON Prompt. Be sure to pause the video and study it. Here too, we see a clear description of the frame, which contains a description of the scene and emotional tone in the scene description and mood keys. Camera movement and lighting are also set as separate parameters, Camera Movement and Lighting, which is important for video production, as well as the overall style of the generated video in the style key value. In this prompt, you can also quickly change only one element without rewriting the entire prompt. For example, set the same scene, but in a different style or with different lighting. If you use neural networks that generate music, then Jon Prompting can also be useful here. Let's look at an example of such a prompt for creating music. This approach is especially convenient if you work with generating tracks for video, podcasts, or media. It allows you to set genres, moods, control the musical instruments used, and specify references so that the AI captures the right vibe. Individual parameters and prompt length depend on the platform. For example, SO has a text request length limitation, and you cannot write a long prompt. Keep this in mind when composing requests. And yes, in general, all of this can also be specified in a regular text request. However, Jon clearly structures the prompt so that neural networks better understand the essence of your idea. But what if you find it difficult to format prompts in the JSON style? For this, we can use our good old chat. You just need to explain to him that you will write your prompt to him, and he should reformat it into JSON for generating the desired content on different platforms. Then it will be enough to describe to chat GPT what you want to get in the end, and he will rewrite your prompt in JSON format, which can be used anywhere. So, let's summarize. Jon Promping is not just a way to beautifully format a request. It is a tool that helps to think structurally, communicate with AI effectively, and achieve desired results faster. This is especially relevant if you work in the media, design, video marketing, or education fields. Although these fields should not be limited. And since JON prompt allows you to get the desired result faster, you save both time and tokens. Be sure to try applying this approach to your work, and you will feel the difference. Share the results of your experiments in the comments. And what to do with this video, you already know. See you.