{"id":58243,"date":"2026-10-17T09:46:19","date_gmt":"2026-10-17T01:46:19","guid":{"rendered":"https:\/\/www.1ai.net\/?p=58243"},"modified":"2026-09-20T15:24:47","modified_gmt":"2026-09-20T07:24:47","slug":"%e6%8f%90%e7%a4%ba%e8%af%8d%e5%88%9b%e4%bd%9c%e7%ac%ac%e5%85%ad%e5%8d%81%e4%b9%9d%e8%8a%82%ef%bc%9a%e6%8a%bd%e5%8d%a1%e5%88%b0%e5%8f%af%e6%8e%a7%e6%8b%8d%e7%89%87%ef%bc%8cseedance2-0","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/58243.html","title":{"rendered":"Section 69 of the AI tip: What exactly is the Seedance tip? \"Spacing Cards\" to Controlable Films, Seedance 2.0 Typists"},"content":{"rendered":"

Use sameAI Video<\/a>Models, people get short films of stable people and consistent lenses, and people get short clips of fine, but irrelevant images\u3002<\/p>\n

Often the problem is not just the ability of the model, nor the lack of sufficient adjectives, but rather that the hint does not tell the model: Who is this, where she is, where the camera starts, what happens next, and where the mirror is to stop\u3002<\/p>\n

A CONTROLLABLE AI VIDEO ALERT CAN BE DIVIDED INTO TWO LAYERS\u3002<\/p>\n

The first level is a global set-up and is responsible for maintaining the identity and quality of the whole work<\/p>\n

The second level is a single-photo command, which is responsible for making the current seconds truly visible. This, together with the end-to-end link between the lenses, gives the opportunity to form a longer narrative\u3002<\/p>\n

This course will explain the reusable writings from the six areas of asset preparation, hint structure, time axis, splitting, continuity and inspection methods\u3002<\/p>\n

\"513b9724jln6hn014xd000v90jp\"<\/p>\n

I. Preparation of assets and then a reminder<\/strong><\/p>\n

\u201cAssets\u201d are not complex documents that are required by a professional team. It refers to all reference materials that need to remain unchanged in multiple videos\u3002<\/p>\n

At least four role views should be prepared for short person-type films, and a clear look reference should be given when there are fixed clothing, hairdressing, weapons or products\u3002<\/p>\n

Scene-type films can also complement space layouts, main colours, materials and key props\u3002<\/p>\n

What about the assets? It's a course to make your own assets\u3002<\/p>\n

Section 67: How does an AI asset stabilize? https:\/\/www.1ai.net\/58213.html<\/p>\n

\"d03a3240jln6i6012nd000v90jp\"<\/p>\n

Level 1: Lock the whole film with a global set<\/strong><\/p>\n

The global set is a common rule for each paragraph. It is not responsible for telling the whole story, but it has established the following:<\/p>\n

Identity, clothing and props<\/p>\n

The scenery age and spatial style<\/p>\n

Image material, light and color<\/p>\n

(a) General emotions<\/p>\n

Changes expressly prohibited\u3002<\/p>\n

\"0aadb77fj00tln6js012nd000v90jp\"<\/p>\n

The following structure may be consulted:<\/p>\n

[photographic parameters and graphic senses] Cooke Panchro\/i Classic Prime Lines, with 1\/4 or 1\/8 Black Pro-Mist radium and focal segment at the front of the camera: 35mm and 50mm focal points are used more frequently in the room than in the day-to-day use of 35mm and 50mm focal areas, with the optical circle controlled between T2.0 and T2.8. Kodak Portra 400 film color. It's called Teal & Warm Amber, no music, no subtitles. Movie machine, camera, color, reference film, etc
\nThe future is in doubt, calm, oppressive, clinical. The opening scene, like a judicial deposition, is gradually creating a feeling of anxiety that \u201cthe system is creating a person\u201d\u3002
\nFig. 1 Young Asian girl. In the dark red red red rags, white knee socks, black Mary Jane shoes, dark red Chinese thongs. She was just plasticized and left with a slight water vapour on her body and clothing surface. (Personal description)
\nFigure 2 Young Asian boys, long hair in black, cold head. Dark blue business leisure shirt or dark business shirt, light long pants (person description)
\nThe 3rd scenario is the anchor
\n[Colourboard] Color tone with colourboard as a whole. (Colourboard anchor)<\/p>\n

\"7175d853jln6k1014d000v90jp\"<\/p>\n

If the work is to be produced in sub-sections, this part can be compressed into a fixed prefix and placed in front of each of the hints. Compression is not the deletion of key rules, but rather the deletion of repetitions and the retention of information that can be observed and checked\u3002<\/p>\n

III. Second floor: conversion of the director ' s thought lens to a single-photo command<\/strong><\/p>\n

It's set to solve the \"what's the whole picture look like\" and the \"how does this mirror work?\"\u3002<\/p>\n

At least three things will have to be said:<\/p>\n

Where the camera starts, who's the main character on the picture, what's the main character doing\u3002<\/p>\n

But it's just the bottom of the second floor, and the upper limit depends entirely on your director's mind lens\u3002<\/p>\n

\"bba0dfa6j00tln6kh0188000v90jp\"<\/p>\n

The more specific the director's spectroscopy, the more complete the second level can be. It determines not only what the person does, but also how the image, the position, the movement route, the performance rhythm, the scene, the slot, the camera movement, the sound and the different images connect. The message itself does not produce director design in a vacuum, but only translates an already established schedule into a language that a video model can execute\u3002<\/p>\n

i've created a skiller of these tips, so as long as they're ready for the director's specs and asset maps of people, scenes, props, etc., we can hand them both together<\/p>\n

Asset-driven Seedance 2.0 Hint Skill<\/p>\n

 <\/p>\n

# Asset-Drive Video Induction Skill<\/p>\n

## Skill Name
\nAsset-Drive-Ai video alert automatically extrapolates Skill<\/p>\n

## Skill positioning
\nThis Skill applies to a user who has completed the assets of a person, scene, prop, palette, style reference, front frame, etc., and uses a simple image to automatically extrapolate the full film message suitable for the AI Video Generation Model\u3002<\/p>\n

Users do not need to write complex tips themselves\u3002
\nThe user needs only to enter a similar type:<\/p>\n

\u2013 The hostess discovered the presence of giants in the desert tribe
\n\u2013 The man rises from the ground and looks to the far sky
\n\u2013 The young girl put the talisman behind his father's back and the father was pushed to the monster
\n\u2013 The monks walk in peace among the chaos
\n\u2013 The camera went from prop to the whole tribe
\n\u2013 The beast stomps on the house and the crowd runs away
\n\u2013 The character picked up the bag and suddenly realized who he was<\/p>\n

You need to automatically extrapolate:<\/p>\n

\u2013 What kind of camera should be used
\n\u2013 Where does the camera start
\n\u2013 How does the camera move
\n\u2013 How the subject is going to enter
\n\u2013 How the move happens in stages
\n\u2013 How assets should be called
\n\u2013 What about the vision in front of the painting
\n\u2013 How can the sound be layered
\n\u2013 Where to write to die, to prevent models from going off
\n\u2013 Final output of a video message that can be directly reproduced<\/p>\n

—<\/p>\n

## Skill Core objective
\nThe goal of Ben Skill is not to allow users to learn complex hints, but to allow AI to take on the work of directors, photographers, spectrographs and hint engineers\u3002<\/p>\n

Users only provide \"image intent\" and you expand it to a full shot\u3002<\/p>\n

The final message should:<\/p>\n

1. Clear access to assets already produced by users
\n2. Automatically complete camera language
\n3. Automatically inducing the course of action
\n4. Automatically complete sound design
\n5. Automatically join model constraints
\n6. Coherence of roles, scenes and proxies
\n7. Video alerts that generate a strong film sense, continuity and enforceability<\/p>\n

—<\/p>\n

## DEFAULT WORKSTREAM<\/p>\n

When users turn on this Skill, you need to work according to the following steps as long as the user enters a simple image:<\/p>\n

#1T#1 STEP 1: IDENTIFICATION OF USER ASSETS
\nFirst, the type of asset that the user may currently own is judged and is explicitly called in the hint\u3002<\/p>\n

Common assets include:<\/p>\n

\u2013 Personal assets: role make-up, four-view, face, costume, action reference
\n\u2013 Site assets: big scenes, overviews, 9 palace scenes, local scenes
\n\u2013 Policing assets: white prop, six view, material reference
\n\u2013 Monster assets: monster make-up, scale, local structure
\n\u2013 Palette assets: main colour, auxiliary colour, light colour, environment colour
\n\u2013 Style assets: photography, photo, texture
\n\u2013 First-end frame assets: video starting frame, video ending frame<\/p>\n

If the asset is not clearly listed by the user, do not follow up. You can write with placeholders:<\/p>\n

\"Role asset: Make-up photo of the woman.\"
\n\"Scene Assets: Desert Nations\"
\n[Property Asset: Red Flour]
\n\u266a Monster asset, giant crustacean \u266a
\n[Colourboard assets: tan, wine, warm ash, brass and green]<\/p>\n

—<\/p>\n

#1T#2 STEP 2: JUDGE THE IMAGE TYPE
\nDepending on the user's brief description, it automatically determines which lens type it belongs to\u3002<\/p>\n

Common types are as follows:<\/p>\n

#1T1T1T#1
\nUser profiles show giants, monsters, dragons, mechanical boats, huge buildings, feelings of oppression, distant appearances, out-of-the-ground and stomping on buildings\u3002<\/p>\n

Priority use:
\n\u2013 Extremely low
\n\u2013 On camera
\n\u2013 Local drawings
\n\u2013 Scale Reference
\n\u2013 Low frequency
\n\u2013 Small crowd contrast<\/p>\n

#1T1T1T#2 CRISIS PREGNITION
\nThe user's description shows: hearing sound, discovering abnormalities, suddenly looking up, above-ground vibrations, moving far away, and the danger of role detection\u3002<\/p>\n

Priority use:
\n\u2013 Fixed camera
\n\u2013 Medium proximity or close-up
\n\u2013 Lifetime action interrupted
\n\u2013 Sound first
\n\u2013 Micro-expression reaction
\n\u2013 The subject of disaster is not fully present<\/p>\n

#1T1T1T#3FOCUS FIELD
\nIn the user's description, there's a calm coming, faceless, confused, high-powered, monk, female master's sense of oppression, anti-fascist comedy\u3002<\/p>\n

Priority use:
\n\u2013 Mid-long focus retreating and filming
\n\u2013 Centred
\n\u2013 Time and time
\n\u2013 Paradigms
\n\u2013 Subject stability
\n\u2013 Small action to avoid interference<\/p>\n

#1T1T1T#4 MOTION ERUPTION
\nThe user's description appears: stickers, impactes, explosions, knives, jumps, blows, pushes to the enemy, energy explodes, fights\u3002<\/p>\n

Priority use:
\n\u2013 Move slowly on the upgrade
\n\u2013 Side-shot or over shoulder
\n\u2013 Action beat break
\n\u2013 Contact clear
\n\u2013 Energy spreads from a little
\n\u2013 Impact clear<\/p>\n

#1T1T1T#5T5 SPACE DISPLAY
\nUser profiles appear: displaying the entire battlefield, seeing the whole tribe, shooting up, looking down, messing up, multiple role positions\u3002<\/p>\n

Priority use:
\n\u2013 General
\n\u2013 Shake your arms
\n\u2013 crane up
\n\u2013dolly back
\n\u2013 The old-middle-middle-middle-middle-middle-middle-middle-middle-middle-middle-middle-middle-middle-mingled vision
\n\u2013 Clear space relations<\/p>\n

#1T1T1T#6 SUSPENDED
\nThe user's description shows up from behind the shades, the curtains are removed, the prop is removed, the crowd is seen, from the black field, the fabric is blocked\u3002<\/p>\n

Priority use:
\n\u2013 Hide the revelations
\n\u2013 Images start from local points
\n\u2013 arc dolly
\n-slow push in
\n\u2013 Sound from far and near
\n\u2013 Colour or material match<\/p>\n

#1T1T##7 EMOTIONAL DIALOGUE
\nIn the user's description, two people spoke, found identity, questioned, surprised, silent, looked at and whispered\u3002<\/p>\n

Priority use:
\n\u2013 Medium Nearview
\n\u2013 Fight back
\n\u2013 Eyes match clips
\n\u2013 Lower light
\n\u2013 Micro-expressions
\n\u2013 The lines close to the lens<\/p>\n

—<\/p>\n

#1T#3 STEP: AUTOMATICALLY SELECT LENS POLICY
\nDo not simply expand user descriptions into images\u3002
\nYou have to judge the best shot strategy\u3002<\/p>\n

The process of analysis can be done without the output of the hint, but internal compliance is required:<\/p>\n

\u2013 If the user wants to shock, use low altitude, local giant, scale reference
\n\u2013 If the user wants an advanced sense, use restraint, slow rhythm, light change
\n\u2013 If the user has to be nervous, use the sound, fix the camera, and the details are abnormal
\n\u2013 If the user wants to hit, use action to keep up, get in contact, get out, get out
\n\u2013 If the user wants to be ridiculous, with the subject stable, the background disorder, the rhythm contrasts
\n\u2013 IF THE USER WANTS A NARRATIVE, USE HARD CUT, VISUAL MATCHING, VOICE CONTINUATION<\/p>\n

—<\/p>\n

## AUTOMATIC EXTRAPOLATION RULE<\/p>\n

### RULE I: USER'S SENTENCE SHOULD BE EXTENDED TO A FULL SHOT
\nUser input:<\/p>\n

\"The hostess discovered the presence of giants far away.\"<\/p>\n

Can't just write:<\/p>\n

The hostess stood in the desert, and there were giants far away, and the images were shocking\u3002<\/p>\n

It should be expanded automatically to:<\/p>\n

\u2013 What is the hostess doing
\n\u2013 What did she hear first
\n\u2013 Which props are vibrating first
\n\u2013 How does she stop
\n\u2013 How does she look up
\n\u2013 How does the camera turn from her to far away
\n\u2013 How can a giant show only its parts
\n- With what reference
\n\u2013 How does the background respond
\n\u2013 How to sound from low frequency to explosion
\n\u2013 Which assets must be consistent<\/p>\n

—<\/p>\n

#1T# RULE II: ASSETS MUST BE \"LOCKED\"
\nWhen creating a hint, the user ' s asset must be clearly preserved\u3002<\/p>\n

Recommended formulation:<\/p>\n

strictly refer to and maintain the looks, age perception, hair, clothing, makeup, body ratio and quality of persons in [role asset: xx], do not change role identity, do not replace clothing and do not create new roles\u3002
\nstrict reference to the spatial structure, architectural style, landscape, colour and light direction of [situation asset:xx]\u3002
\nstrict reference to the shape, material, colour, texture and size ratio of [protocol asset:xx]\u3002
\nstrict reference to [colourboard asset:xx] controls the overall colour and does not add high saturation colours that do not belong to the world view\u3002<\/p>\n

—<\/p>\n

#%s# RULE III: PHRASINGS MUST AUTOMATICALLY COMPLETE THE \"PURMEDIATE MEDIUM VISION\"
\nEvery shot has a spatial dimension\u3002<\/p>\n

Recommended structure:<\/p>\n

The visual vision is [the element of the outlook], the medium scenario is [the main action], and the vision is [the landscape is deep\/the crowd\/the giant\/the sky\/the architecture]\u3002
\nThe subject is located in the [location] of the image, which is about [proportionally], and the background is formed by [mog\/smoky\/light\/dark\/fire] layers\u3002<\/p>\n

—<\/p>\n

#%s# RULE 4: ACTIONS SHALL BE SPLIT INTO BEATS
\nEvery move must be broken down into a continuous process\u3002<\/p>\n

Standard action chain:<\/p>\n

The start-up state, the trigger, the main reaction, the initiative, the key contact, the result, the outbreak, the after-wave feedback<\/p>\n

For example:<\/p>\n

At the beginning, the hostess looked down and looked at the talisman in her hand\u3002
\nIt's coming from a remote low-frequency vibration, and the edges of the amphibians are shaking\u3002
\nHer hand stopped and she was breathing briefly\u3002
\nShe looked up to the top right\u3002
\nThe camera follows her vision to a distance\u3002
\nThe horns of the giants slowly detect from the dust\u3002
\nThe crowd was shaken back, and the tent cloth and dust were shaking\u3002<\/p>\n

—<\/p>\n

#%s# RULE V: GIANT MUST WRITE REFERENCE
\nIf there's a giant in the picture, it's not allowed to say \"great.\"\u3002
\nA clear reference must be written\u3002<\/p>\n

Available references:<\/p>\n

\u2013 Human
\n\u2013 Tents
\n\u2013 Houses
\n\u2013 A camel
\n\u2013 Vehicles
\n\u2013 The walls
\n\u2013 Trees
\n\u2013 Wall
\n\u2013 Mechanical ship
\n\u2013 Props<\/p>\n

Recommended sentence:<\/p>\n

The giants are so large that the images cannot be fully contained that only [locally] appear and the rest of the body extends outside the frame. By contrast, the [reference] is already several times greater than the [reference], and the prospective human being is as small as an ant at its feet\u3002<\/p>\n

—<\/p>\n

### RULE 6: SOUND MUST BE GENERATED AUTOMATICALLY
\nA voice must be written regardless of whether it is spoken by the user\u3002<\/p>\n

The sound is divided into four layers:<\/p>\n

1. Bottom environment sound
\nWinds, night worms, low-frequency ringing, distant crowds, flames, rain, sand and dust<\/p>\n

2. Mid-level action
\nFootsteps, run, impact, drag, roll, weapons collision, monster movement<\/p>\n

3. Proximity
\nClothes rubbing, metal touch, hair dripping, breathing, finger pinching, sandstone scratching<\/p>\n

4. Popping point
\nThe giants roar, the earth roars, the buildings collapse, the energy explodes, the crust breaks<\/p>\n

Recommended sentence:<\/p>\n

At the sound level, the bottom is covered with [the sound of the environment], the middle is penetrating [the sound of action], the periphery is clear [the sound of detail], the critical blast point is added [the sound of explosion], the lines are close to the lens and are clearly floating above the sound field\u3002<\/p>\n

—<\/p>\n

### RULE VII: AUTOMATICALLY ADD TO THE ANTI-RUN BIAS LIMIT
\nComplex lenses have to be bound\u3002<\/p>\n

Common constraints:<\/p>\n

\u2013 Keeping role assets consistent
\n\u2013 Keep the dress the same
\n\u2013 Don't change a person's face
\n- Don't change the scenery
\n\u2013 No new irrelevant
\n\u2013 The subject does not drift
\n\u2013 The camera does not spin meaninglessly
\n\u2013 Don't suddenly change the scenery
\n\u2013 The giants don't show up completely
\n\u2013 Do not deform the props
\n\u2013 Do not create subtitles
\n\u2013 No background music
\n\u2013 No modern items
\n\u2013 No buildings or costumes that don't fit the worldview
\n\u2013 All jammers must enter from the edge of the picture, not in the air<\/p>\n

—<\/p>\n

## Output format<\/p>\n

When the user gives a brief description of the image, the output is in the following format\u3002<\/p>\n

#1T# IMAGE TYPE JUDGEMENT
\nOne sentence describes the type of image, such as:<\/p>\n

This picture is for \"Crisis Perception + Masses Up\" and it's not about shooting monsters from the start, it's about pulling people's attention away with sound and detail\u3002<\/p>\n

#1T# CALL ON ASSETS
\nLists the assets that need to be called in for this reminder\u3002<\/p>\n

Format:<\/p>\n

\u2013 Role asset: [insert role]
\n\u2013 scene asset: [insert scene]
\n\u2013 Props assets: [insert props]
\n\u2013 Monster asset:
\n\u2013 Palette asset: [insert palette]
\n\u2013 Style asset: [insert style]<\/p>\n

If the user does not provide a specific asset name, the placeholder is used\u3002<\/p>\n

#1T1T# DIRECTLY DUPLICABLE VIDEO PLUGIN
\nOutput complete hint\u3002
\nThis part must be complete and replicable without explanation\u3002<\/p>\n

—<\/p>\n

## FINAL TIP STRUCTURE<\/p>\n

Video alerts that can be copied directly must use the following structure:<\/p>\n

\"Asset Consistency Constraints\"
\nStrict reference is made to and maintains the profile, hair, clothing, makeup, body ratio and quality of the person in the [role asset]; strict reference is made to the spatial structure, architectural style, landscape, light orientation and colour in the [situation asset]; strict reference is made to the shape, material, texture and size ratio in the [prototype asset]; and strict reference is made to the [padboard asset] control the overall colour. Do not change role identity, do not replace clothing, do not add irrelevant persons or modern objects. Only environmental sound, action sound and line sound are generated and do not match background music\u3002<\/p>\n

\"The camera set.\"
\n[Specific] [Specific], [Specific]. The camera starts with the [starting position] and moves in the [simulation mode] to the [motion direction] and stops at the [deposition point]\u3002<\/p>\n

Image Structure
\nThe visual outlook is [the element of the outlook], the medium view is [the subject and the action] and the vision is [the environment\/population\/megastat\/sky\/construction]. The subject is in the image [place], which is about [proportion]. Background is formed by [sane depth\/dust\/mog\/night\/fire] layers\u3002<\/p>\n

Action beat
\nAt the beginning of the event. Then the [trigger] appears, the subject [reaction action], then [active action], in [critical moment] [contact\/turn\/discovery]. And then it happens, and afterward it gets back\u3002<\/p>\n

Light and Mass
\nLights come from [direction], light with [photo] [main part], and form on [face\/metallic\/ cloth\/dust\/mog] [shading\/reflective\/shading level]\u3002<\/p>\n

\u266a Emotions and performances
\nEmotions are expressed through the eyes, the breath, the lips, the hands, the body pauses, without exaggerating\u3002<\/p>\n

Sound Design
\nThe bottom is covered with [environmental sound], the middle is penetrating [acting sound], the periphery is visible [detail sound], the critical blast point is added [blasting sound], and the lines are floating above the sound field [sonic\/distance\/emotion]\u3002<\/p>\n

\"Model constraints\"
\nThis mirror must hold several lines from the beginning to the end: [Treatment 1]; [Treatment 2]; [Treatment 3]; [Treatment 4]; [Treatment 5]\u3002<\/p>\n

—<\/p>\n

## AUTOMATICALLY GENERATED TEMPLATES FOR ONE PHRASING<\/p>\n

When the user does not require multiple lenses, the default generation of a full single lens hint\u3002<\/p>\n

The structure is as follows:<\/p>\n

Strict reference and consistency between [role asset], [situation asset], [protocol asset] and [colour-board asset] does not change role appearance, clothing, morphology and landscape structure. Generates only the sound of the environment and the action, which does not match the background music\u3002<\/p>\n

[Specific] [Specific], [Specific], lenses moving from [Specific] to [Frame]. The visual vision is [foreground], the medium view is [subject], the vision is [environment\/incident] and the subject is [place]. At the start [initial state], then [trigger] appears, the subject [microresponse] occurs, then [active] occurs, forming [results] in [critical moment]. Lights enter from [direction], light [main part], so that [material\/dust\/fog\/metallic\/matrix] produces real dark and dark layers. The bottom of the sound is the voice of the environment, the middle is the sound of action, the proximate view is the sound of detail and the blast point is the sound of an explosion. This mirror must remain stable [subject\/scenario\/spectrum\/assets\/actions] without adding irrelevant elements and without changing the asset set\u3002<\/p>\n

—<\/p>\n

## AUTOGENERATION TEMPLATE<\/p>\n

When the user says \"Do a set of lens\" \"Do a video\" \"Do a slide\" \"15 seconds\" \"Multiple lens\" it automatically generates 3 to 6 lenses\u3002<\/p>\n

EACH SHOT MUST BE SEPARATED BY THE HARD CUT\u3002<\/p>\n

The structure is as follows:<\/p>\n

Camera one: [Building Space\/Crisis Perceptions\/Main State]
\nHARD CUT
\nCamera 2:
\nHARD CUT
\nCamera three:
\nHARD CUT
\nCamera 4: [Explosion\/discovery\/shock result]
\nHARD CUT
\nCamera 5:<\/p>\n

Every shot has to be resubmitted:
\n\u2013 A farewell
\n\u2013 Flight slot
\n\u2013 Structure
\n\u2013 Subject Position
\n\u2013 Action beat
\n\u2013 Sound<\/p>\n

—<\/p>\n

## AUTOSELECT CAMERA TEMPLATE BASED ON IMAGE DESCRIPTION<\/p>\n

### USER INPUT: ROLE SEES DISASTER IN A DISTANCE
\nUse:
\nCrisis Perception + Vision Match + Vision Display<\/p>\n

Default extrapolation:
\nThe role is first filmed in its routine or static state, followed by an abnormally interrupted voice, and the role is up, and the lens reveals the disaster in the distance\u3002<\/p>\n

—<\/p>\n

### USER INPUT: BEAST APPEARANCE
\nUse:
\nVery low slot + local drawing + scale reference<\/p>\n

Default extrapolation:
\nNot just the giant, but only the feet, horns, backs, tails or shadows. A comparison is made with crowds, houses, vehicles, tents\u3002<\/p>\n

—<\/p>\n

#1T#INPUT: THE ROLE HAS PASSED THROUGH CHAOS
\nUse:
\nMedium long focus retreat + centered + background disorder<\/p>\n

Default extrapolation:
\nRoles remain stable, context is chaotic, jammers are emptied from the edges, and the role is avoided in very small actions\u3002<\/p>\n

—<\/p>\n

#%s# USER INPUT: ROLE PROP
\nUse:
\nProps + Hand action + Contact point outbreak<\/p>\n

Default extrapolation:
\nScratch the details of the props, then the hand contact, and finally the results of the props. The consistency of props ' assets was emphasized\u3002<\/p>\n

—<\/p>\n

### USER INPUT: FIGHTING A COLLISION
\nUse:
\nRaise slow action + side beat + action beat + impact feedback<\/p>\n

Default extrapolation:
\nTake the power, then contact, then explode, then impact, and then after\u3002<\/p>\n

—<\/p>\n

### USER INPUT: DISPLAY FULL SCENE
\nUse:
\nShake the arm up + Space reveals + the front-middle vision expands<\/p>\n

Default extrapolation:
\nThe footage starts in local areas and is gradually being raised to allow people, buildings, props, vision events to enter the picture\u3002<\/p>\n

—<\/p>\n

### USER INPUT: CHARACTER FINDS AN IDENTITY OR SECRET
\nUse:
\nEyes match clips + props + face reactions<\/p>\n

Default extrapolation:
\nFirst, the role vision changes, then the props or clues are cut, then the role micro-expressions are removed\u3002<\/p>\n

—<\/p>\n

## EXAMPLE INPUT AND OUTPUT METHOD<\/p>\n

#1T# EXAMPLE INPUT
\nI have the masters, the desert tribes, the red feathers, the giant crustaceans. The picture is that the lady found the beasts from afar\u3002<\/p>\n

### SHOULD BE EXPORTED
\nImage type judgement:
\nThis picture is for \"Crisis Perception + Masses on the Ground\" and focuses first on the sound and detail abnormally on the female owner, and then on the discovery of the giant by low-level and scale reference\u3002<\/p>\n

Asset call:
\n\u2013 Role assets: make-up photos of women
\n\u2013 Site assets: desert tribal landscape
\n\u2013 Pedestrian assets: red truffles
\n\u2013 Monster assets: giant crustaceans
\n\u2013 Palette assets: desert tan, wine red, warm ash, copper green<\/p>\n

Video alerts that can be copied directly:
\nAnd then you're going to output the full hint\u3002<\/p>\n

—<\/p>\n

## DON'T DO ANYTHING<\/p>\n

1. Do not allow the user to complete the complex camera itself\u3002
\n2. Do not just output keywords\u3002
\n3. Not just a picture\u3002
\n4. Do not ignore user assets\u3002
\n5. Do not change role costumes and appearances\u3002
\n6. Do not let the giants appear in their entirety\u3002
\n7. Do not let action take place in vain\u3002
\n8. Do not ignore voices\u3002
\nDon't let the camera wander\u3002
\n10. Do not be written in a general narrative\u3002<\/p>\n

—<\/p>\n

## ULTIMATE PERFORMANCE STANDARD
\nWhen the user enters a simple image, you complete it like a director:<\/p>\n

\u2013 Why is this shot taking like this
\n\u2013 Where should the camera be
\n\u2013 What should the subject take
\n\u2013 How should assets be called
\n\u2013 How does that happen
\n\u2013 Where did the audience first look
\n\u2013 How sound guides emotions
\n\u2013 What must be locked to death<\/p>\n

THE FINAL OUTPUT MUST BE A HINT THAT CAN BE COPIED DIRECTLY INTO THE AI VIDEO MODEL\u3002<\/p><\/blockquote>\n

This Skill will directly generate a set of available Seedance 2.0 tips based on visual information in the slides\u3002<\/p>\n

let me show you how this skill works\u3002<\/p>\n

upload skill first, upload your assets and your spectroscopy, click to send\u3002<\/p>\n

\"a192af6dj00tln6lc00esd000v9000tlp\"<\/p>\n

\"f56434c1jln6li007id000my01c8p\"<\/p>\n

AI GENERATED HINTS<\/p>\n

In this process, the director's think lens decides what to do, how to do it, the asset map locks out what people and the world look like, and Skyll organizes them into enforceable tips\u3002<\/p>\n

Case: What is the real strength of Life<\/p>\n

A REPRESENTATIVE EXAMPLE OF THIS IS THE \"LIFE IS LONGER THAN DREAM\" FILM SERIES OF AI BY THE SHIVERING CREATORS, LIFE IN LIFE. THE WORK INTEGRATES THE MYTHS OF THE EAST, THE DOWRY, THE MECHANICAL CIVILIZATION AND THE SABER VISION INTO THE SAME WORLD AND CREATES A RECOGNIZABLE WORK BY FULL NARRATIVES, PROMINENT ROLES AND CONTINUOUS LENSES\u3002<\/p>\n

\"ef7bf6daj00tln6ly00xad000v9000jip\"<\/p>\n

The true strength of the author is his mind, his lens and his aesthetic skills. The director's mind determines what the story says, what the audience sees first; the spectroscope ability determines how the person comes out, how the camera moves, how the information is revealed; and the aesthetic ability determines whether colour, light, image, sound and visual elements belong to the same world. Indications and Skills are important, but they are located at the executive level and are responsible for handing over these creative decisions to the model rather than to the creator\u3002<\/p>\n

\"a41fc868jln6md01azd000v90jp\"<\/p>\n

IV: Three notes for the second level of hint<\/strong><\/p>\n

1. Write in camera order, do not preset action time<\/strong><\/p>\n

Before it is generated, we cannot determine exactly how many seconds it takes to lift a hand, turn a turn or push a mirror. The pace is affected by the speed of the movement of the person, the way the model is implemented and the length of the final segment. Therefore, it is not recommended that the actions be rigidly divided into periods of \u201c0-4 seconds, 4-8 seconds\u201d in advance. The seemingly accurate number of seconds may, on the contrary, cause movement to be compressed, slowed or crowded out\u3002<\/p>\n

\"7195c87bj00tln6mq012yd000v90jp\"<\/p>\n

What really needs to be written is action sequence and lens. If the picture remains the same, it shows directly where the camera starts, how it moves, who it always follows, and where it ends. For example: \u201cThe mid-vision begins with the person's right rear, the camera follows the person into the room; the person is parked behind the table, the camera is pushed slowly to her right hand, and finally stops at the moment when the finger is about to touch the file bag.\u201d<\/p>\n

If the middle needs to be hard-cut, the front-and-forward image is written in two paragraphs, with the word \u201chard-cut\u201d clearly written between the two paragraphs. The model thus knows that the previous shot is over and that the latter is to start with a new view or slot\u3002<\/p>\n

If it's not hard, write down specific lenses in the hint. For example, if you need to move closer, write that the camera is slowly approaching the person's face from the middle view; if you need a shift of focus, you need to move the focus from the prospect paper bag to the person in the back; rather than just write \u201cthe natural transition of the lens\u201d, because \u201cnatural\u201d has no enforceable graphic standard\u3002<\/p>\n

This writing does not prejudge the time of the model, but rather gives the model a clear sequence. When the action is simple, the picture can naturally slow down; when the action is larger, the mirror should continue to be removed instead of imposing a number of seconds for each action\u3002<\/p>\n

2. Complex mirrors to be removed so that each mirror has only one focus<\/strong><\/p>\n

Many of the hints look professional, but they are written at one time in terms of proximity, around, up and down, focus, focus shift and rapid switching. Even if the models recognize these words, they do not necessarily keep the person ' s movement, space direction and image stable at the same time. A more reliable way to do this is to do it by narrative: this mirror shows action, the next one shows clues and the next one shows the reaction of the person\u3002<\/p>\n

\"369f152aj00tln6n5015wd000v90jp\"<\/p>\n

THE SPLIT SHOULD BE SUBJECT TO THE DIRECTOR'S THOUGHT LENS. IF THE CAMERA WAS DESIGNED TO BE A CONTINUOUS LONG SHOT, THE COMPLETE COURSE OF THE CAMERA WAS WRITTEN, AND NO SPLITS WERE NECESSARY. HARD CUT (HARD CUT) IS A DIRECT SWITCH; DOLLY IN (TRACK APPROACH) IS A CAMERA MOVING TO THE SUBJECT; RACK FOCUS (FOCUS SHIFT) IS A MOVE FROM ONE OBJECT TO ANOTHER. WHEN THE TERM IS FIRST USED, IT IS ADVISABLE THAT THE CHINESE IMAGE BE WRITTEN AT THE SAME TIME, SO AS TO AVOID A STACK OF NAMES\u3002<\/p>\n

\u00a0The bottom of the lens is the director's dispatch lens<\/strong><\/p>\n

Locking the end image is not an added technique to create, but the bottom is actually director movement and spectrometry. The director's Scheduling Spectroscope determines what's in each image, where the person is standing, how it moves, where it looks, where the props are and where the camera is taken. As long as the content of the adjacent images is well organized, the lenses generated by this lens will naturally be relevant and the connection will be smoother\u3002<\/p>\n

An executive director's movement spectroscope, which should contain five types of information: spatial relations, including the location of persons, tables, chairs, doors and windows and key props; movement of persons, including positions, movement routes, turn-ups, rises and stops; performance relationships, including action, expression, vision and reaction between persons; lens movements, including scenes, slots, lens movements and video focus; and lens relations, including continuous lenses, rigid, back-to-back or other means of connection\u3002<\/p>\n

The logic of character dialogue. The spectroscope provides for the female master to sit on the left side of the table, the male master to sit on the right side, two face-to-face; the male master rises and looks to the female master. Next, you can cut the man's view, and you can put him on the left side of the picture, and then you can hit the woman, and you can put her on the right side of the picture. This is not a single \u201clock\u201d that ends in a single end, but is already written in the director's control spectroscope in terms of the person's position, the direction of the view and the relationship between the counterplay\u3002<\/p>\n

Continuity can be summed up in one sentence: the image of the adjacent lens is reasonable and the connection to the lens is reasonable. The same action, vision, proprietary position and movement direction shall be subject to the space relationship determined above when changing the position or position. The clearer the director's control spectroscope, the easier it is to write a single-scenario hint, and the less it will be necessary to collage it by means of remediation\u3002<\/p>\n

V. How to check and modify after generation<\/strong><\/p>\n

The completion of the hint does not mean that there must be one success. Model versions, the quality of reference maps, the length of time generated and the capacity of the platform influence the results. At least five checks are made after generation: the identity of the person is consistent; the movement is carried out by the director; the camera is not carrying out the main movement; the key props are retained; and the adjacent image is compatible with the space and action relationship specified in the lens\u3002<\/p>\n

In case of a problem with the generation of results, there is a need to continue to adjust and recreate the hints, and to contrast the picture over and over again with the director's thinking, until the desired effect is achieved. Each round is as specific as possible: changes in the person ' s face strengthens the asset map and appearance limits; without the implementation of the mirror, the irrelevant description is deleted and the camera movement is rewritten; and, when the lens is unconnected, the person ' s position, movement, vision and props are checked back in the adjacent lens\u3002<\/p>\n

To recognize the boundaries of capabilities. The reference map does not guarantee 100 per cent consistency; too many roles, fast physical contact, complex props and many simultaneous actions remain prone to error. The asset-driven Seedance 2.0 tipword, Skill, can directly generate a set of available hints, provided that the director's spectrometer and asset map are available; however, it implements the director's design, and lens creativity, movement logic and final trade-offs still depend on the creator\u3002<\/p>","protected":false},"excerpt":{"rendered":"

THE SAME IS TRUE OF THE AI VIDEO MODEL, WHERE PEOPLE GET SHORT CLIPS THAT ARE STABLE AND CONSISTENT, AND PEOPLE GET SHORT CLIPS THAT ARE FINE BUT IRRELEVANT. TOO OFTEN THE PROBLEM IS NOT JUST THE ABILITY OF THE MODEL, NOR THE LACK OF ADJECTIVES, BUT THE HINTS DO NOT TELL THE MODEL WHO IT IS, WHERE SHE IS, WHERE SHE STARTS, WHERE THE CAMERA STARTS, WHAT HAPPENS NEXT, AND WHERE THE MIRROR STOPS. A CONTROLLABLE AI VIDEO ALERT CAN BE DIVIDED INTO TWO LAYERS. THE FIRST LEVEL IS A GLOBAL SET-UP, WHICH IS RESPONSIBLE FOR MAINTAINING THE IDENTITY AND QUALITY OF THE ENTIRE WORK; AND THE SECOND LEVEL IS A SINGLE-PHOTO COMMAND, WHICH IS RESPONSIBLE FOR MAKING A VISIBLE MOVE IN THE CURRENT SECONDS. THIS, TOGETHER WITH THE END-TO-END LINK BETWEEN THE LENSES, GIVES THE OPPORTUNITY TO FORM A LONGER NARRATIVE. THE CURRICULUM WILL BE SPLIT FROM ASSET PREPARATION, HINT STRUCTURE, TIME AXIS, MIRRORS, CONTINUITY AND INSPECTION METHODS<\/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-58243","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\/58243","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=58243"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/58243\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=58243"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=58243"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=58243"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=58243"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}