T8mars/minimax-h3-prompt-skill-T8 / skills/direct-street-interview-video
direct-street-interview-video
VerifiedDesign natural vertical street-interview and walk-and-talk video prompts for MiniMax H3 or Seedance 2.0. Use when a user wants spontaneous documentary energy, a referenced recurring person, first-person handheld follow, concise dialogue, moving daylight, believable street parallax, or a consent-led interaction that must fit 4–15 seconds without unreachable beats.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 5.7 KB |
| agents/openai.yaml | config | 326 B |
| references/h3-template.md | markdown | 3.7 KB |
| references/seedance-template.md | markdown | 2.5 KB |
| references/source-analysis.md | markdown | 2.7 KB |
| references/summary.md | markdown | 3.4 KB |
| references/template.md | markdown | 5.2 KB |
references/h3-template.md
View on GitHub# MiniMax H3 template
Use the Base form when there are no generation-time references. Use Ref2VA only when the user actually supplies an image or video that must control identity, wardrobe, scene, motion, or sound.
## Ref2VA skeleton
```text
subject_definitions:
<Subject 1> is the on-camera [role] whose identity, facial structure, hairstyle, body proportions, and [wardrobe anchor] come from <Picture 1>. Preserve natural skin texture and the same outfit throughout.
<Picture 1> is the identity and wardrobe reference for <Subject 1>; it is not a mandatory first frame.
summary:
[reference generation] Create a [TARGET_DURATION]-second [ASPECT_RATIO] documentary street interaction in which <Subject 1> [one-sentence social premise and resolved state].
retention_analysis:
<Subject 1>: fully_preserved — preserve identity, proportions, hairstyle, and wardrobe while allowing natural walking, speaking, gaze, and gestures.
<Picture 1>: attribute_transfer — transfer identity and wardrobe only; rebuild the street, route, framing, lighting, and action as an original scene.
detailed_description:
Natural vertical street-documentary image quality, realistic skin and fabric response, coherent background parallax, and restrained handheld motion. No unsupported resolution claim.
[Shot 1] A first-person camera [backpedals / side-walks] at [framing] on an original [street setting]. <Subject 1> begins [opening action] at [opening distance]. [Moving daylight behavior] crosses both the subject and the street; [foreground/background anchors] shift with continuous parallax. The off-screen interviewer (S1) says <d>[Chinese][SHORT LINE 1]</d>. <Subject 1> (S2) looks toward the lens, [micro-performance], and replies <d>[Chinese][SHORT LINE 2]</d>. Their steps remain synchronized with the route, with realistic foot contact and small step-linked camera sway.
[Optional Shot 2] At [MM:SS.mmm below the target duration], [only if a genuine event boundary is needed]. <Subject 1> [visible relationship-changing action] while the camera [one motivated reframe]. (S1) says <d>[Chinese][OPTIONAL SHORT LINE 3]</d>. (S2) [gesture or gaze response] and replies <d>[Chinese][OPTIONAL SHORT LINE 4]</d>. The action resolves as [permission / pose / reveal / route choice / natural release], held long enough to read before the video ends. Do not begin a new setup at the endpoint.
overall_soundscape:
Continuous location ambience from [street sound family], soft synchronized footsteps, fabric movement, and only the physical sounds caused by visible actions. Dialogue is not repeated here. No abrupt ambience reset across the optional cut.
non_diegetic_music:
N/A
```
Delete the optional shot when the scene is a single continuous event. Define only labels that correspond to real generation-time inputs, and include exactly one retention line per independent label.
## T2VA skeleton
```text
integrated_multimodal_description: [Shot 1] A [TARGET_DURATION]-second [ASPECT_RATIO] original street-documentary scene. [Stable subject appearance], [social premise], [one route], [one primary camera follow], [performance progression], [moving daylight and parallax], [approved dialogue with speaker IDs and <d> blocks], and [readable resolved state].
overall_soundscape: [Continuous location ambience, synchronized footsteps and action foley, no repeated dialogue.]
non_diegetic_music: N/A
```
## H3 audit
- First shot has no timestamp; later timestamps increase and remain below the target duration.
- All prose is English except approved dialogue inside language-tagged `<d>` blocks.
- Dialogue appears only in the visual timeline.
- No Seedance `镜头N`, braces, or `@图片N` syntax appears.
- No research video is declared as an input unless the user separately supplies it for generation.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation). | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK). | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |