{"id":58358,"date":"2026-10-20T10:45:39","date_gmt":"2026-10-20T02:45:39","guid":{"rendered":"https:\/\/www.1ai.net\/?p=58358"},"modified":"2026-09-20T15:34:05","modified_gmt":"2026-09-20T07:34:05","slug":"ai%e6%8f%90%e7%a4%ba%e8%af%8d%e5%88%9b%e4%bd%9c%e7%ac%ac%e5%85%ab%e5%8d%81%e8%8a%82%ef%bc%9a%e4%b8%ba%e4%bb%80%e4%b9%88ai%e4%ba%ba%e7%89%a9%e6%80%bb%e6%9c%89%e5%81%87%e4%ba%ba%e6%84%9f","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/58358.html","title":{"rendered":"Section 80 of the AI tip: How do you write the perfect tip with Codex? Let's stop hand-to-hand and let Ai learn to write for you"},"content":{"rendered":"

A lot of people make itAI Video<\/a>NOW, THE FIRST THING TO DO IS TO OPEN A LARGE MODEL, TO DESCRIBE THE THOUGHTS IN YOUR HEAD BRIEFLY, AND THEN LET AI DIRECTLY GENERATE A HINT. IT SEEMS VERY EFFICIENT, BUT WHEN IT REALLY COMES TO BEING CREATED, IT IS OFTEN DISCOVERED THAT PEOPLE ARE NOT DOING THE RIGHT THING, THAT THE SPECIAL EFFECTS ARE NOT IN THE RIGHT SHAPE, AND THAT THE CAMERA RHYTHM IS NOT THE SAME AS THEY THOUGHT, AND THAT THEY CAN ONLY CHANGE THEIR TIPS AND REPEAT THEIR CARDS\u3002<\/p>\n

THE PROBLEM DOES NOT NECESSARILY ARISE FROM MODELLING CAPABILITIES, BUT FROM THE VERY BEGINNING WE GAVE AI TOO MUCH \"SPECIFY SPACE\". IF AI DOESN'T EVEN KNOW WHAT YOU WANT, THEN EVEN IF IT WRITES A FEW HUNDRED WORDS OR EVEN A THOUSAND WORDS, IT'S JUST HELPING YOU OUT WITH YOUR OWN UNDERSTANDING\u3002<\/p>\n

In this section, we change our thinking: first let AI learn what to write and then hand over to it the whole idea; then, when there is a need, it does not rush to produce; then it turns back; and then it goes back to Seedance, and it changes on the basis of actual results\u3002<\/p>\n

What is really needed is not a one-size-fits-all word, but a set of phrases that can be used repeatedly\u3002<\/p>\n

FIRST, DON'T RUSH TO WRITE A HINT. LET AI LEARN WHAT YOU WANT TO DO FIRST<\/strong><\/p>\n

WE USED TO DO AN AI VIDEO, WE USED TO HAVE A HABIT: WHEN AN IDEA CAME UP IN OUR HEAD, WE OPENED THE AI MODEL IMMEDIATELY, AND THEN WE ENTERED A VIDEO TIP TO HELP ME WRITE A FILM FEELING, TO HELP ME WRITE A SPECIAL EFFECT TIP FOR A CHARACTER TO TRANSFORM, AND I WANTED A VERY POWERFUL ENERGY BURST EFFECT\u3002<\/p>\n

\"8bafd83bj00tlner006nd000v90efp\"<\/p>\n

an ai-direct hint<\/p>\n

OF COURSE, AI CAN GIVE THE ANSWER IMMEDIATELY, AND USUALLY IT'S COMPLETE. THERE MAY BE DESCRIPTIONS OF \u201cMIDDLE SHALLOW DEPTH, CLEAR SUBJECT, NATURAL DEFUNCT BACKGROUND, CLEAR FRONT AND BACK\u201d \u201cENERGY ERUPTIONS, PARTICLE AGGREGATION, LIGHT SWEEPS, MATERIAL RE-ENGINEERING, FRAGMENTATION\u201d. LOOKING AT THIS HINT ALONE, IT SEEMS LIKE THERE IS NO PROBLEM, BUT WHEN IT REALLY COMES UP, IT IS OFTEN FOUND THAT THE EFFECT IS NOT WHAT WE IMAGINE\u3002<\/p>\n

The same is true of the transformation of people, who want to see luminous cracks on their hands and then gradually spread over their arms; who want to be surrounded by flames; and who want to turn clothes into mechanical armour. They can all be called \u201cdiversions\u201d, but what really controls the results is where change begins, how sequenced, how long it lasts, and what it turns into\u3002<\/p>\n

So first we have to change a habit:<\/p>\n

DON'T LET A.I. WRITE YOU A FINAL TIP, LET IT UNDERSTAND WHAT YOU WANT TO DO\u3002<\/p>\n

first, go to higgsiled\/any shared website to find public cases<\/p>\n

here you can go straight to the higgsiled for a public case close to your target\u3002<\/p>\n

\"31b797edj00tlnesi00old000v9000fqp\"<\/p>\n

\"adbe1cbej00tlnesv00gyd000v9000fmp\"<\/p>\n

This is a crucial step, because it is best for us not to show AI a final video, but to provide it with the results of the case together with the original hint. The core method of the original video-text case was also to go to the higgsiled first to look for a few public cases that had close effects, and then hand over the results to AI together with the original hint\u3002<\/p>\n

Why must both be provided<\/strong><\/p>\n

IF IT'S ONLY IN FILM, AI CAN CERTAINLY ANALYZE WHAT'S GOING ON, WHAT'S GOING ON, HOW THE CAMERA MOVED, WHAT'S GOING ON, BUT IT DOESN'T KNOW EXACTLY WHAT LANGUAGE THE CREATOR USED TO CONTROL THE RESULTS\u3002<\/p>\n

Once the original phrase has been included, it can establish a very important correspondence\u3002<\/p>\n

It can judge which part of the hint is in control of the person, which part affects the rhythm of the action, how the effects are described, how the lens movement is expressed, and which limits are preventing the person from deforming, changing the lens or changing the scene\u3002<\/p>\n

WE'RE NOT JUST LETTING AI READ IT, BUT WE'RE JUST LETTING IT LEARN:<\/p>\n

What exactly is the hint behind this kind of work\u3002<\/p>\n

If conditions permit, three to five similar cases can be found at a time\u3002<\/p>\n

For example, we're going to create a \u201cperson release trick\u201d, so we can search for a series of application cases\u3002<\/p>\n

\"9b654f41jlnetb00zvd000v9000v8p\"<\/p>\n

These cases do not need to be fully consistent with their final ideas, as long as they belong to similar visual types\u3002<\/p>\n

WHEN MULTIPLE CASES APPEAR AT THE SAME TIME, AI IS MORE LIKELY TO FIND COMMON GROUND, RATHER THAN MECHANICALLY IMITATING A HINT\u3002<\/p>\n

WE CAN JUST TELL AI:<\/p>\n

I'm going to give you a couple of open AI video cases that you found from Higgsiled, each with the final results and the corresponding original hints. Do not now generate new hints. Please analyse how each set of hints controls people, actions, lenses, effects, time sequences and constraints, taking into account the image, and summarize the structure and expression of the hints that are common to multiple cases. At this stage, only analysis and learning are undertaken, and my final hint is not generated\u3002<\/p>\n

The most important here is:<\/p>\n

\"128616500jlnetj004rd000qj00lzp\"<\/p>\n

The first cycle is only learning, not creating\u3002<\/p>\n

A LOT OF PEOPLE ARE USED TO HANDING OVER ALL TASKS ONCE AND FOR ALL TO AI: \"HELP ME ANALYZE THESE CASES AND WRITE A NEW ONE AS I WANT.\"<\/p>\n

THIS SAVES ONE DIALOGUE, BUT AI TENDS TO MOVE QUICKLY INTO THE LATER GENERATION TASK, AND THE ANALYSIS OF CASES IS SIMPLE\u3002<\/p>\n

A more stable way would be for it to focus on how this effect should be described\u3002<\/p>\n

WE MOVE TO THE SECOND STEP WHEN AI HAS A SUFFICIENTLY CLEAR REFERENCE\u3002<\/p>\n

II. DON'T START WITH PROFESSIONAL WORDS AND TELL AI YOUR TRUE THOUGHTS<\/strong><\/p>\n

After the reference, it's our own idea\u3002<\/p>\n

But this step should not be rushed into writing \u201cindicators\u201d\u3002<\/p>\n

Because there are many people here who fall into another wrong zone: in order to make themselves professional, they begin to struggle to find photography terms, special effects names and complex visual adjectives\u3002<\/p>\n

The image in the original mind may be clear:<\/p>\n

The person normally moves forward, and a few seconds later the right hand starts to have blue cracks, which slowly spread to the shoulder, and the lens gets a little closer when the energy becomes stronger\u3002<\/p>\n

But when it really started, it was easy to become:<\/p>\n

\"Technology blue energy changes, film-level visual effects, shock particle effects, epic photons.\"<\/p>\n

It seems more professional, but the information that is really useful is less\u3002<\/p>\n

BECAUSE THE PRECEDING PARAGRAPH CONTAINS A VERY CLEAR SEQUENCE: WHAT THE PERSON IS DOING AT FIRST, WHERE THE CHANGE TAKES PLACE, HOW IT GOES, WHEN THE LENS CHANGES. THE LATTER PARAGRAPH GAVE AI ONLY A FEW BROAD DIRECTIONS, AND MUCH OF THE REMAINING DETAIL REQUIRED ITS OWN GUESS\u3002<\/p>\n

So what really should be pursued is not \u201chow professional is the language\u201d, but:<\/p>\n

The information is not complete\u3002<\/p>\n

IF THE NEED IS MORE COMPLEX, I WOULD PREFER TO USE THE VOICE INPUT FUNCTION OF THE AI MODEL DIRECTLY. THIS IS ALSO EMPHASIZED IN THE ORIGINAL VIDEO ARTICLE: IT IS EASIER FOR A VOICE TO TELL THE WHOLE PICTURE OF A DISORDER THAN A TEXT ENTRY, AND AI CAN FURTHER HELP US ORGANIZE OUR NEEDS\u3002<\/p>\n

For example, you can just say:<\/p>\n

You make me feel like this, that's why this guy went ahead in the first place, and didn't have to do anything. And then for about two seconds, there was a sudden movement on the ground, as if it was going to crack, and then it burst out of the ground, not a little bit, and it was the kind of ice crystal with a sense of aggression. And then, when these crystals explode, they come up to the top or attack it directly, with a sense of surprise. The camera begins with a normal follow-up to the person, and when the ground starts to crack and the crystal is coming out, then lean a little further forward, but not in the person's face. The face of the person, the main costumes and the background are kept as far as possible, and the whole process is not sudden-cut, and the image follows\u3002<\/p>\n

\"51fe4699jlnetx004ad000om00bwp\"<\/p>\n

This is not like the \u201cstandard phrase\u201d we normally see, or even some repetition\u3002<\/p>\n

BUT FOR AI, THIS EXPRESSION IS VERY VALUABLE\u3002<\/p>\n

Because the initial state of the person, the changing position, the sequence of development, the motion of the camera, the conditions of preservation and the prohibited content are all covered\u3002<\/p>\n

This is also why complex needs need not be over-compressed\u3002<\/p>\n

YOU CAN DESCRIBE IT IN ITS OWN WAY OF THINKING, IF YOU CAN JUST TELL THE REAL MESSAGE. THE PERSON IS IN WHAT STATE HE OR SHE WOULD BE, IN WHAT ORDER HE OR SHE WOULD CHANGE, AND IN WHAT ORDER HE OR SHE WOULD EVENTUALLY WISH TO STOP; IF THE CAMERA IS CLEARLY REQUIRED, HE OR SHE CAN FOLLOW IT OR REMAIN FIXED. IN THE CASE OF MULTI-PERSON INTERACTIONS, THERE IS A NEED TO CLARIFY THE SEQUENCE AND PLACE OF THE PERSON. IF THE PERSON ' S APPEARANCE, CLOTHING, SCENE OR SOME KEY MOVE MUST BE MAINTAINED, IT SHOULD ALSO BE SHARED WITH AI\u3002<\/p>\n

It can be spoken very well\u3002<\/p>\n

AFTER WE'RE DONE, LET'S GET A.I<\/p>\n

Our natural language expressions tend to be repetitive or may be confusing, so do not immediately ask for a final hint\u3002<\/p>\n

LET AI ORGANIZE OUR NEEDS INTO A CLEAR IMPLEMENTATION STATEMENT\u3002<\/p>\n

For example:<\/p>\n

This is my description of the natural language of the target video, where there may be duplication or confusion. Don't help me to generate the final hints or add new ideas. Please organize the person, scene, movement, effect change, lens movement, time sequence, what needs to be kept and what is forbidden to come, on the basis of my original needs. If there is a conflict, please also indicate separately\u3002<\/p>\n

\"470c4725j00tlneu9003wd000rq00klp\"<\/p>\n

When it is sorted out, a scattered description becomes a clear video demand\u3002<\/p>\n

This step has actually helped us to accomplish a very difficult thing in the creation of traditional hints:<\/p>\n

WE NO LONGER NEED TO TRANSLATE IMAGES FROM OUR OWN MINDS INTO \u201cAI LANGUAGE\u201d\u3002<\/p>\n

WE ARE ONLY RESPONSIBLE FOR EXPRESSING OUR IDEAS AS FULLY AS POSSIBLE, SORTING AND CONVERTING THEM TO AI\u3002<\/p>\n

IF YOU NEED IT, DON'T CREATE IT RIGHT AWAY<\/strong><\/p>\n

Once the needs are sorted out, the next step will be for many people to go directly into the Seedance generation\u3002<\/p>\n

But for a slightly more complex video, I suggest that one more step be made here:<\/p>\n

LET'S START WITH WHAT AI NEEDS TO PREPARE TO REVERSE THE TARGET\u3002<\/p>\n

It's a very important part of the whole work stream\u3002<\/p>\n

THE IMPORTANCE OF THE LIST OF MATERIAL WAS ALSO HIGHLIGHTED IN THE ORIGINAL VIDEO, AS THE MORE COMPLETE THE MATERIAL IS PREPARED, THE LESS CONTENT AI NEEDS TO SPECULATE ABOUT ITSELF\u3002<\/p>\n

The current multi-mode video model no longer requires all information to be written\u3002<\/p>\n

We can provide personal reference maps, scene maps, action videos, lens references, special effects cases, etc. directly\u3002<\/p>\n

\"0736ae0aj00tlneus00fid000v9000f2p\"<\/p>\n

For example, if a person ' s appearance is not well described, a reference chart of the person is given; if an action is not known, a reference video of the action is given; if it is not known what the effect is, it can be given directly to a similar case\u3002<\/p>\n

From this perspective, the role of the phrase has changed\u3002<\/p>\n

In the past, we wanted a hint to give the whole picture from zero\u3002<\/p>\n

The more rational way is now:<\/p>\n

The material is responsible for providing specific information and the hint is responsible for defining the relationship between such information\u3002<\/p>\n

LET AI DECIDE WHAT IT REALLY NEEDS<\/strong><\/p>\n

HAND OVER TO AI THE NEED TO ORGANIZE THE FRONT, SO THAT IT DOES NOT GENERATE THE FINAL HINTS, BUT RATHER REVERSES THE MATERIAL ACCORDING TO THE TARGET\u3002<\/p>\n

For example:<\/p>\n

BASED ON CURRENTLY IDENTIFIED VIDEO REQUIREMENTS, NO FINAL HINTS SHOULD BE WRITTEN. PLEASE REVERSE WHAT MATERIAL IS REQUIRED TO COMPLETE THIS VIDEO AND CLASSIFY IT ACCORDING TO WHAT MUST BE PREPARED, RECOMMENDED AND NOT PREPARED. EACH OF THE MATERIALS NEEDS TO INDICATE WHAT IT IS RESPONSIBLE FOR CONTROLLING, WHAT IT SHOULD NOT BE CONSULTED ON, AND WHERE AI IS MOST LIKELY TO MAKE MISTAKES WITHOUT IT. FINALLY, GIVE ME A SET OF MATERIALS THAT CAN BE AS STREAMLINED AS POSSIBLE\u3002<\/p>\n

\"7f9ffff0j00tlnev6002td000k00mzp\"<\/p>\n

AL GAVE ME THE REQUIRED MATERIAL<\/p>\n

After analysis, we may have different material such as character reference maps, scene reference maps, action videos and effects reference\u3002<\/p>\n

Here's a special note:<\/p>\n

The material is not as good as it is, but as clear as it is\u3002<\/p>\n

What really matters is:<\/p>\n

It is desirable that each resource should have only one clear primary responsibility\u3002
\n
\nONCE THE MATERIAL IS READY, LET AI WRITE THE FINAL VIDEO ALERT<\/strong><\/p>\n

When the list of materials is ready, we actually enter the process of creating the hint\u3002<\/p>\n

THIS IS WHEN WE CAN CONTINUE TO TELL AI:<\/p>\n

Now that the needs and materials have been fully identified, please create for me the final AI video alert based on the public case writing you have learned before. The hints need to be combined with all the material that I have prepared, and to make clear what each material is responsible for controlling and not allowing different references to interact. Also, in the order in which the video actually took place, the movement of the person, the presence of the monster, the explosion of the ground, the appearance of the crystal, the attack of the crystal against the monster and the change of the lens were clearly written. Do not add to the new circumstances that I did not propose, and do not create a multiplicity of adjectives to appear professional. Ultimately, only one video message in Chinese can be sent directly to Feedance\u3002<\/p>\n

Note: The final hint is to put the material in context<\/p>\n

ONCE THE MATERIAL IS READY, LET AI GENERATE THE FINAL VIDEO ALERT\u3002<\/p>\n

At this point in time, the introduction begins by identifying the use of each material\u3002<\/p>\n

For example:<\/p>\n

Reference is made to figure 1 on the appearance of the person, hair and clothing; reference is made to the scene set in figure 2 on the image of the monster in figure 3<\/p>\n

\"59323b7cjlnevt009d000qa00ncp\"<\/p>\n

Finally, it adds to the conditions that must be maintained, such as the stability of the face of the person, the unchanging structure of the scene, the non-involved role, the special effects that do not shield the person ' s face, and the effects that follow should occur in advance\u3002<\/p>\n

The model is no longer a multiplicity of descriptions\u3002<\/p>\n

It will be clearer:<\/p>\n

Which material is responsible for what happens first, what happens afterwards, and what cannot be changed\u3002
\n
\nThe first generation is just a test, and the real hint optimization starts after the result<\/strong><\/p>\n

Once the material has been prepared and the final hint is made, it will be able to enter the Feedance generation\u3002<\/p>\n

But it's better to change your mind:<\/p>\n

The first generation is not the final answer, but the first test\u3002<\/p>\n

For more complex videos, it is difficult to require that the model fully meet expectations for the first time\u3002<\/p>\n

We can begin to judge very specifically:<\/p>\n

Whether or not the actions are correct, where the effects are, when the lens moves, if the scene changes, and at what point the whole video begins to deviate from the target\u3002<\/p>\n

1. RE-TRANSMIT GENERATION RESULTS TO AI<\/p>\n

IF THE RESULTS ARE NOT SATISFACTORY, THE RESULTING VIDEO CAN BE RE-ASSIGNED TO THE AI MODEL TO ALLOW IT TO BE RE-ANALYSED IN THE CONTEXT OF THE ORIGINAL NEEDS AND THE PRECEDING HINTS. THIS IS THE LAST STEP IN THE ORIGINAL VIDEO: IF THE RESULT IS NOT SATISFACTORY, THE RESULTING VIDEO IS RESUBMITTED TO AI FOR MODIFICATION; IF A PROBLEM IS DIFFICULT TO DESCRIBE BY WORD, IT IS POSSIBLE TO DIRECTLY INTERCEPT AND MARK THE WRONG IMAGE\u3002<\/p>\n

\"a63c7752jlnewa000cd000p1005qp\"<\/p>\n

I don't know how to describe the problem<\/p>\n

There is also a very common situation in practice:<\/p>\n

You can see at once what's wrong, but you don't know how to describe it\u3002<\/p>\n

It may be that the person's arm has a strange structure, that the effect is slightly different, that an element behind the person should not appear, or that there has been an inexplicable distortion in a particular area\u3002<\/p>\n

There is no need to force yourself to look for professional terms at such times\u3002<\/p>\n

TAKE THE WRONG IMAGE AND MARK THE PROBLEM POSITION WITH A CIRCLE, ARROW OR SIMPLE TEXT AND GIVE IT TO AI\u3002<\/p>\n

\"264f5258j00tlnewh002hd000p00008bp\"<\/p>\n

VISUAL PROBLEMS ARE TOLD DIRECTLY TO AI, AND IN MANY CASES THEY ARE MORE ACCURATE\u3002<\/p>\n

Summarizing: The true almighty is not a hint, but a full stream of work<\/strong><\/p>\n

At the heart of this section is, in fact, the transformation of \u201cdirectly written hints\u201d into a more stable process\u3002<\/p>\n

First go to higgsiled to find a close public case, let AI learn the relationship between the results and the original hint; then tell AI what it really thinks and let it sort you out. It then identifies the respective references for characters, scenes, actions and effects on the basis of the list of demand-backed materials, and then produces an official video alert and gives it to Seedance\u3002<\/p>\n

IF THE FIRST GENERATION IS NOT DESIRABLE, THEN THE RESULTS ARE RESUBMITTED TO AI FOR ANALYSIS. PROBLEMS THAT ARE DIFFICULT TO DESCRIBE CAN BE TRANSCRIBED DIRECTLY, WITH ONLY THE PROBLEMATIC PARTS MODIFIED AND WITHOUT THE NEED TO REWRITE THE ENTIRE HINT EACH TIME\u3002<\/p>\n

The set of processes can be summarized as follows:<\/p>\n

higgsiled looks for cases of AI learning, input ideas, sorting needs, inverting materials, generating video alerts, and Seedance generating feedback changes\u3002<\/p>\n

What is really worth understanding is not a \u201cone-size-fits-all message\u201d, but the ability to find reference, to express needs, to plan, and to adapt on the basis of results\u3002<\/p>","protected":false},"excerpt":{"rendered":"

ONE OF THE FIRST THINGS THAT MANY PEOPLE DO WHEN THEY MAKE AN AI VIDEO IS TO OPEN UP A BIG MODEL, DESCRIBE THE IDEAS IN THEIR HEAD BRIEFLY, AND THEN LET AI DIRECTLY GENERATE A HINT. IT SEEMS VERY EFFICIENT, BUT WHEN IT REALLY COMES TO BEING CREATED, IT IS OFTEN DISCOVERED THAT PEOPLE ARE NOT DOING THE RIGHT THING, THAT THE SPECIAL EFFECTS ARE NOT IN THE RIGHT SHAPE, AND THAT THE CAMERA RHYTHM IS NOT THE SAME AS THEY THOUGHT, AND THAT THEY CAN ONLY CHANGE THEIR TIPS AND REPEAT THEIR CARDS. THE PROBLEM DOES NOT NECESSARILY ARISE FROM MODELLING CAPABILITIES, BUT FROM THE VERY BEGINNING WE GAVE AI TOO MUCH \"SPECIFY SPACE\". IF AI DOESN'T EVEN KNOW WHAT YOU WANT, THEN EVEN IF IT WRITES A FEW HUNDRED WORDS OR EVEN A THOUSAND WORDS, IT'S JUST GOING TO HELP YOU OUT ON YOUR OWN UNDERSTANDING. IN THIS SECTION, LET'S CHANGE OUR THINKING: LET'S LET'S LET'S LEARN WHAT TO WRITE FIRST, AND THEN GIVE IT THE WHOLE IDEA; YES<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[149,144],"tags":[802,3149,956,5321,192,6481],"collection":[],"class_list":{"0":"post-58358","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"hentry","6":"category-jiaocheng","7":"category-baike","8":"tag-ai","12":"tag-prompt","13":"tag-6481"},"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/58358","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=58358"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/58358\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=58358"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=58358"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=58358"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=58358"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}