1. Skill Objective
Based on the 1–10 real travel photos and real travel locationsuploaded by the user, automatically complete:
- Read and analyze the subject, scene, sense of time, color tone, depth of field, and movable elements of each photo;
- Without destroying the real content of the photos, re-plan a more reasonable storyboard order;
- Generate a set of unified visual master template prefixes；
- For each photo, output one independently generatable first-frame image-to-video prompt；
- By default, design at 3–6 seconds per shot, and prioritize recommending that users generate each shot independently, then proceed to post-production editing.
This Skill is intended for users with no AI video experience. The core goal is not to showcase complex camera movement, but to let real travel photos naturally "come alive," forming a unified, restrained, cinematic cultural tourism promotional film.
2. Default Input and Recommended Specifications
Supported Input
- Number of photos: 1–10
- Recommended number: 3–6
- Location: real travel locations such as cities, scenic areas, streets, buildings, coastlines, natural landscapes, etc.
- Optional information:
- Desired total duration
- Desired per-shot duration
- Landscape / portrait
- Whether there are people
- Whether natural ambient sound is needed
- User-specified photo order or shots that must be retained
Default Values
When the user provides no additional instructions:
- Aspect ratio: 16:9
- Per-shot duration: 4 seconds
- Reasonable range: 3–6 seconds
- Generate one video independently for each photo
- No subtitles and no background music throughout
- By default, use hard cuts or natural match cuts in post-production, and do not require the video model to generate complex transitions
If the user uploads 7–10 photos at once, still process all photos normally, but note:3–6 photos usually make it easier to form a short cultural tourism film with concentrated rhythm and visual unity.
3. First Principles
3.1 The first frame determines "what the image is"; the Prompt only determines "how the image moves"
Do not repeatedly describe the large amount of architecture, clothing, scenery, materials, and spatial details already clearly present in the photo.
The prompt should focus only on writing:
- How the camera moves;
- How the subject naturally continues its action;
- Which environmental elements can move slightly;
- Which visual information must remain stable;
- What overall emotion is desired.
3.2 Content first, camera movement as support
The core of a cultural tourism promotional film is: Location, people, the breath of life, the natural environment, and spatial beauty.
Camera movement is only a tool serving the image.
It is forbidden to forcibly add, just to "look high-end":
- Large orbit;
- Fast push-pull;
- Whip pan;
- Drone dive;
- Hitchcock zoom;
- Multiple consecutive focus pulls;
- Complex speed ramping;
- Camera movement that does not match the spatial relationship with the first frame.
3.3 One primary cinematographic intent per shot
In principle, each shot should only choose:
- 1 primary camera movement
- At most 1 auxiliary technique
For example:
- Slow tracking shot + natural character movement
- Static shot + focus pull
- Slow lateral move + foreground-background parallax
Do not stack multiple techniques within a 3–6 second shot.
3.4 Leave room for the video model to perform
Unless the action rhythm must be controlled, avoid second-by-second choreography.
Prefer using:
- "Slow"
- "Natural"
- "Slight"
- "Stable"
- "After about 2 seconds"
Avoid large amounts of:
- Precise angles
- Displacement percentages
- Lens millimeter numbers
- Complex physical parameters
- Frame-by-frame action descriptions
3.5 Continuity of real locations and people
Real locations
- Prefer using the city, scenic spot, or neighborhood names explicitly provided by the user as the cultural tourism context;
- Design movement only based on content actually visible in the photos; do not invent new landmarks, buildings, folk elements, or weather just because the place name is known;
- If the user only provides a city name and the photos cannot confirm a specific scenic spot, do not proactively guess a specific landmark;
- Place names are mainly used to unify the cultural tourism context and mood, and do not serve the function of redesigning the scene.
Same person
When the same person appears in multiple photos:
- Prefer arranging related shots more continuously;
- Keep the person's identity, clothing, and overall temperament consistent;
- Make action directions form a natural continuation as much as possible;
- Do not force the creation of new actions without basis in the photos for the sake of a "continuous storyline."
4. Unified visual master generation rules
The unified master must be based on the user's real photos, rather than forcibly transforming all scenes into the same filter.
Default master
{Location} cinematic cultural tourism promotional film, {aspect ratio}, ARRI Alexa + 35mm film look, real natural light and physical reflections, fine natural film grain, low-saturation cinematic color grading, premium, restrained, realistic. Uniformly continue the relationships of people, architecture, space, weather, and light in the input photos, slow and stable camera movement, producing only subtle natural environmental motion, no obvious scene reconstruction, no subject deformation, no exaggerated CG feel. No subtitles and no background music throughout.
Color adaptation rules
When most photos themselves suit golden hour, sunset, blue hour night scenes, or city lights, you may add:
Warm golden highlights, cyan-blue shadows
When the photos are mainly:
- Noon natural landscapes
- Snow scenes
- Forests
- Rainy cities
- Interiors
- Highly saturated local cultural scenes
Do not mechanically apply "warm gold × cyan-blue."
Instead:
Using the original photo's natural color tones as the base, unify with low-saturation cinematic color grading and soft highlight layering.
The role of the visual master is to "unify," not to "override."
5. Photo Analysis
Do only the necessary analysis for each photo, internally judging the following information:
- Subject: people / architecture / streets / coast / mountains and rivers / indoors / food / animals / night scenes, etc.
- Shot size: close-up / medium shot / full shot / extreme long shot
- Visual focus: the position in the frame that truly needs to be viewed
- Spatial layers: whether there is obvious foreground, midground, background
- Sense of time: early morning / daytime / golden hour / blue hour / night / unclear
- Dominant color tone
- Content that can move naturally
- Whether any person is already in the middle of an action
- Whether there are elements that can be used for match cuts
- Whether it is suitable for special cinematography techniques
Analysis is only for decision-making; do not write a lengthy image analysis report in the final answer.
6. Storyboard Ordering Rules
The order in which the user uploads photos by default does not equal the final edit order。
Unless the user explicitly requests keeping the original order, proactively provide a more reasonable recommended order.
Prioritize the following four types of continuity
A. Temporal continuity
When photos have obvious temporal changes, prioritize a natural timeline:
early morning → daytime → golden hour → sunset → blue hour → night
But do not break apart consecutive shots that clearly belong to the same action, same person, or same space just for the sake of the timeline.
B. Subject continuity
Prioritize placing the following photos adjacent:
- the same person
- the same family / companions
- different angles of the same landmark
- the same street in different shot sizes
- before-and-after relationship of the same action
C. Visual match continuity
Can use:
- similar color tones
- similar compositions
- similar motion directions
- similar lines
- visual elements such as water surfaces / lights / doorways / stone steps / sky
to form simple match cuts.
D. Narrative rhythm
When there is no obvious temporal or character continuity, you can use:
location establishment → entering the space → people / life → core landmark / Hero Shot → relaxed breathing → city / night scene / distant view ending
It is not required that every film fully applies this.
Opening and ending
Prioritize selecting:
- Opening: photos with location recognizability, a sense of spatial establishment, or strong foreground layering
- Ending: photos with a sense of "coming to rest," such as sunset, night scene, extreme long shot, still people, city skyline, sea surface, etc.
7. Single-Shot Duration Selection
Default is 4 seconds, automatically adjusted within the 3–6 second range according to content.
3 seconds
Suitable for:
- simple architecture
- static landmarks
- small push-in
- simple subtle movement of people
- transition shots in fast-paced short videos
4 seconds
Default value, suitable for most travel photos.
5–6 seconds
Suitable for:
- focus pull
- Character transitions from stillness to walking
- Coastal lateral tracking
- Scenery that needs a longer breathing feel
- Time-lapse photography
- Slight slow-motion action
Do not force a static photo into a 6-second complex action just to extend the duration.
8. Camera Movement Selection Library
Architecture / Landmarks
Preferred:
- Static shot
- Slow push-in
- Slow lateral tracking
- Slight lateral parallax
Avoid large orbits around buildings.
Streets / Plazas / Walking People
Preferred:
- Slow follow shot from behind
- Lateral follow shot
- Steady slow push-in
Characters should maintain a realistic walking pace and natural state.
Portraits / Travel Photography
Preferred:
- Static shot
- Extremely slow push-in
- Slight lateral parallax
Keep character actions simple:
- Blinking
- Breathing
- Turning the head
- Slight look back
- Naturally starting to walk
- Hair and clothing swaying gently in the wind
Do not let the character suddenly perform large-scale actions.
Coast / Lakes / Rivers / Open Landscapes
Preferred:
- Slow lateral tracking
- Static camera position
- Very slight push-in
Let:
- Water surface
- Reflections
- Boats
- Clouds
- Plants
carry the main motion.
City Night Views / High Vantage Points
Preferred:
- Slow lateral tracking
- Slight push-in
- Nearly static
Suitable as an ending.
Indoors / Cafes / Hotels / Museums
Preferred:
- Static shot
- Slow push-in
- Small slider lateral tracking
Avoid altering the internal spatial structure out of nowhere.
Food / Handicrafts / Close-up Details
Preferred:
- Static close-up
- Slight sliding
- Slow push-in
Focus on letting steam, hands, and small-scale material changes occur naturally.
9. Special Cinematography Techniques
Special techniques are only used when the photo itself has clear conditions for them.
Rack Focus
Applicable:
- There are clear subjects in both the foreground and background;
- The depth of field layers are obvious;
- The focus shift can convey information.
For example: Foreground stone monument → distant temple
In principle, only one rack focus per shot.
Foreground Blur
Applicable:
- Leaves
- Railings
- Door frames
- Stone monuments
- Street-side objects
When they already form an obvious foreground occlusion.
Do not add foreground out of nowhere.
Time-lapse Photography
Only applicable:
- Clouds
- Sunset
- City lights gradually turning on
- Crowd flow
- Traffic flow
- Scenes that clearly have value in showing changes over time
When there is a close-up of a person or a person is the core subject, do not use time-lapse by default.
Slow motion / slow-motion feel
Only applies to:
- Wind blowing hair
- Skirt hem
- Waves
- Water splashes
- Flock of birds
- Falling objects
- Natural actions with obvious dynamic beauty
Ordinary walking and static buildings do not use slow motion by default.
Parallax
When the photo has clear foreground, midground, and background, a slight lateral movement can create natural parallax.
Do not force it when there is no spatial layering.
10. Rules for People and Environmental Actions
People
Only continue the most natural next action from the first frame.
Prioritize:
- Breathing
- Blinking
- Slight head turn
- Continue walking
- Naturally start walking from stillness
- Looking at the scenery
- Clothing and hair move gently in the wind
Avoid:
- Suddenly looking at the camera
- Suddenly running
- Waving broadly
- Strong acting
- Interactions with no basis in the first frame
Environment
Prioritize giving natural subtle movement to elements already present in the photo:
- Leaves
- Clouds
- Smoke
- Water surface
- Reflections
- Lights
- Fountains
- Boats
- Distant pedestrians
- Pets
Do not add a large number of new elements out of thin air.
Animals
Maintain the original positional logic and behavioral logic.
For example:
- A dog walking naturally and wagging its tail
- A bird flying slightly
- A boat rising and falling with the water surface
Avoid sudden high-speed movement.
11. First-Frame Stability Rules
Each prompt by default requires:
- Start naturally from the input photo;
- Continue the original person's identity, clothing, and pose;
- Maintain the architecture and spatial structure;
- Maintain the main light source direction;
- Do not redesign the scene;
- Do not arbitrarily add people or objects;
- Do not turn real scenic spots into fantasy buildings;
- No obvious deformation of people, buildings, or animals;
- Do not use exaggerated CG effects.
If the photo contains important and clearly readable:
- Place names
- Stone tablets
- Shop signs
- Building plaques
And the text is the core information of the shot, then additionally add:
Keep the original text content and position stable, without garbled characters or deformation.
Do not mechanically add text locking to every shot.
12. Prompt Generation Formula
Use for each storyboard:
Unified visual master + main camera movement + natural subject action + 1–2 environmental micro-movements + necessary stability requirements + emotion
It is recommended to keep the storyboard-specific description within 2–4 sentences。
Do not rewrite the photo as a text-to-image prompt.
Recommended granularity
Good phrasing:
The camera slowly and steadily trucks along the coastal boardwalk. The person and pet naturally continue walking forward, small ripples form on the sea surface, and distant boats gently bob. Overall quiet and relaxed, with a local seaside atmosphere of everyday life.
Not recommended:
The camera moves 2.7 meters to the right along the X-axis at a speed of 1.5%/s, 35mm lens, the person's right foot steps forward at 0.8 seconds...
13. Output Structure
In the end, output only the content the user actually needs to execute.
A. Plan Summary
Brief explanation:
- Number of input photos
- Recommended overall order
- Estimated total duration
- Overall narrative logic
B. Recommended Storyboard Order
Format:
Original Image 3 → Original Image 1 → Original Image 4 → Original Image 2
Just add a brief reason for each image.
C. Unified Visual Master
Output a complete prefix that can be copied.
D. Final Prompt for Each Photo
Format:
Storyboard 01｜Original Image X｜4s
Camera movement: slow truck
{统一视觉母版}
该镜头专属提示词
Output all photos in sequence.
Each one must be able to be copied individually and used directly for first-frame image-to-video, so each one needs to include the unified visual master; writing only "same as above" is not allowed.
E. Brief Generation Suggestions
Fixed reminder:
It is recommended to generate an independent 3–6 second video for each photo separately, and after you are satisfied, proceed to editing in the recommended order; do not let the video model complete the transitions between all photos in one go.
Unless there are special needs, do not output lengthy photography theory.
14. Interaction Rules for Beginners
When the user has already provided:
- Photos
- Travel location
Start the analysis directly; do not continue asking about a large number of professional parameters.
Ask only when the following information would significantly change the result:
- The user clearly needs vertical orientation but the photos are horizontal;
- The user requires a fixed total duration;
- The user requires that the original photo order must be maintained;
- The user requires a specific brand / platform / video model format;
- The user wants to add narration, subtitles, or music.
Otherwise, complete the plan using all default values.
15. Mixed Aspect Ratio Handling
When the input photo ratios are inconsistent:
- By default, still base it on the user's target video aspect ratio;
- Use 16:9 when unspecified;
- If a photo needs substantial cropping to fit, do not forcibly change the subject composition during the video stage;
- The user should be reminded to prioritize completing outpainting or recomposition first, and then do image-to-video.
16. Editing Continuity
This Skill is only responsible for generating an independent video for each photo, but when ordering, post-production editing needs to be considered.
Default priority:
- Hard cut
- Action direction match
- Color tone match
- Composition match
- Lighting match
Do not proactively generate:
- AI Morph
- Scene dissolve
- Portal
- Camera passing through objects
- Complex transition effects
If two photos naturally have:
- The same person's action
- The same direction of movement
- Same landmark, different shot scales
- Similar doorway / sun / water surface / architectural lines
It may be noted in the ordering description that it is suitable for a match cut.
17. Final quality check
Before output, check only the following issues:
- Whether each photo has an independent Prompt;
- Whether each Prompt includes a unified visual master;
- Whether it prioritizes continuing the first frame rather than reconstructing the first frame;
- Whether one shot has only one main photographic intent;
- Whether the camera movement serves the content;
- Whether the character's actions are natural;
- Whether enough room is left for the environment to develop naturally;
- Whether there are unnecessary precise parameters;
- Whether the storyboard ordering has clear continuity;
- Whether it recommends by default generating each shot independently.
If deleting one sentence from a prompt does not affect the expected result, delete it first.