Skillstravel-skill
T
Featured

travel-skill

Plan, write, review, and repair cinematic cultural tourism promotional films created by blending authentic licensed footage with AI-generated shots. Designed for creative briefs for cultural tourism shorts, scene image selection, opening-frame spatial analysis, storyboard scripts, visual content prompts, character actions, director-level camera movements, naturally motivated lighting, transitions and end-frame continuity, Seedance 2.0/Kling/Midjourney prompts, generation result diagnostics, timeline checks, and compliance risk assessments. Particularly suitable for projects requiring realism, vastness, a sense of freedom, and continuity in character-driven narratives.

travel-skill: AI Cultural Tourism Film Director

Skill Overview

travel-skill is an AI film director skill designed for the production of cultural tourism promotional videos. It treats prompts as executable cinematography directions and covers the complete workflow from creative planning, first-frame spatial audits, and storyboard scripting to prompt writing for Seedance 2.0 / Kling / Midjourney, generated-result diagnosis, and compliance risk checks. It is intended for producing cinematic cultural tourism shorts that combine genuinely authorized footage with AI-generated shots.

Applicable Scenarios

  1. Pre-production planning for cultural tourism promotional videos: Outputs creative directions and the main narrative thread, distinguishes between genuinely authorized footage (landforms, snow-capped mountains, lakes, villages, and human-interest documentary footage) and AI-generated shots (fixed-character storylines, difficult-to-reshoot actions, and controlled transitions), and lists risks and missing-footage requirements.
  2. Image-to-video storyboarding and prompt writing: First conducts a spatial audit of the starting-frame image to confirm what is already present in the frame and which elements have visible entry paths. It then outputs shot-by-shot generation packages in the order of “narrative objective → visual content by time segment → shot size and composition → character actions → camera direction → lighting → sound → end frame.”
  3. Generated-result diagnosis and localized repair: Attributes issues such as chaotic shots, character discontinuity, and lighting inconsistencies to one of eight categories—space, action, camera movement, characters, lighting, continuity, technical issues, and risks. It modifies only the specified shots, keeps shot numbers stable, and does not rewrite the entire film without authorization.

Core Functions

  1. First-frame spatial audit: Treats the input starting frame as the spatial reality of the current generated clip. It first analyzes the subjects, entrances and exits, and occluding objects already present in the foreground, midground, and background, then designs character actions and camera movement accordingly. This prevents the generation of grass, roads, doors, windows, or characters that do not exist in the frame.
  2. Director-level camera movement and motivated lighting: First specifies the camera’s starting point, trajectory, speed, subject being followed, and stopping point. Each generation includes only one primary camera movement. Lighting must have an explainable source: establish the ambient exposure first, then shape the subject. When moving between locations, the default approach is to split the sequence into two clips and connect them with a sound bridge or match-on-action transition.
  3. Multi-model prompt adaptation and quality control: Adapts prompts to the respective duration and character limitations of Seedance 2.0, Kling, and Midjourney. It follows a three-round iteration process: first testing the subject’s action and a single camera movement, then adding shot size and stopping point, and finally supplementing focal length and lighting. Quality control is performed with prompt, timeline, and storyboard checklist validation scripts.

Frequently Asked Questions

Who is this skill suitable for?

It is suitable for cultural tourism promotional film directors, cultural tourism marketing and short-video production teams, AI video creators, and project teams that need to combine genuinely authorized footage with AI-generated shots while maintaining realism and continuity in character-driven narratives.

Can it prevent the problem of things “appearing out of nowhere” in AI videos?

This is precisely the skill’s core constraint. Before generation, the starting frame must undergo a spatial audit to confirm what is already present and whether the character’s actions and camera movement can be physically completed within the specified duration. Any object that does not exist in the starting frame and has no visible entry path must not be included in the prompt. This reduces, at the source, the generation of roads, occluding objects, or characters that appear out of nowhere.

Which generation models are supported?

It includes prompt formulas for Seedance 2.0, Kling, and Midjourney. For image-to-video generation, it prioritizes describing “which elements move, how they move, how the camera moves, and where it finally stops.” Seedance 2.0 and Kling support first-to-last-frame transitions; the end frame should be used only when precise endpoint control is genuinely needed. Midjourney prompts are subject to character-count limitations.

How are real footage and AI-generated shots divided? What compliance reminders apply?

By default, the skill assigns landforms, lakes, canyons, villages, animals, and human-interest documentary footage to genuinely authorized material, while AI handles fixed-character storylines, difficult-to-reshoot actions, and controlled transitions. For material authorization, portrait rights, and religious or folk-cultural content, it provides clear risk reminders only; it does not determine whether the user has “obtained authorization.” Authorization must still be confirmed by the project team.

What should be done if the generated result contains chaotic shots?

First remove the second camera-movement track and irrelevant style words rather than continuing to pile on adjectives. Then identify the cause using the eight-category failure directory—spatial contradictions, actions exceeding the duration, confused camera movement, inconsistent characters, and so on. Report only the cause and proposed modification, and keep shot numbers and timecodes stable during localized repairs.