pov-short-film-generator

Create one complete immersive first-person short film from a locked subjective-camera premise, supplied references, POV-specific storyboard, atomic clips, continuity-safe assembly, and final film QC.

SKILL.md

GitHub

POV Short Film Generator

Catalog

  • Display name: POV Short Film Generator
  • Categories: Film, Video
  • Status: Ready
  • Featured: yes
  • Route label: Subjective narrative film composition
  • Tone: purple
  • Short description: Produce a coherent first-person narrative short with POV-specific continuity and final-film QC.

What this skill does

  • Converts a premise into a locked first-person viewpoint contract instead of a third-person camera plan.
  • Builds and generates atomic subjective actions with visual, sound, and continuity anchors.
  • Assembles one finished film and verifies the viewpoint never breaks unintentionally.

How to use

Provide the premise, intended feeling, duration, ratio, supplied references, POV identity/action constraints, visual direction, audio plan, required moments, and forbidden elements.

Example

Create a 20-second 16:9 first-person short film of someone entering an abandoned greenhouse to rescue a trapped bird. The view is always their eyes; show only hands or partial body when needed, with rain on the glass and no third-person face reveal.

Workflow

  1. Lock the final film, protagonist's subjective role, premise, audience feeling, duration, ratio, first-person rules, permitted protagonist fragments, references, style, audio, required beats, and forbidden views. Reject a premise that requires unapproved identity recreation or cannot maintain subjective point of view.
  2. Write a POV film specification: what “I” see, hear, touch, and do; what is visible at the frame edge; where attention moves; and which third-person shots, face reveals, or camera resets are forbidden.
  3. Assign reference roles and build a timed storyboard. Every clip has one subjective action, view direction, visible fragment if any, external object, start/end state, camera/body movement, style anchor, sound evidence, continuity anchors, and acceptance. Continue to the supported generation route without a client-side payment confirmation.
  4. Select one current cataloged model per clip: Kling V3 Omni for complex/motion references, Kling 3.0 for precise frame control, Seedance 2.5 for audiovisual/artistic work, Veo 3.1 for eligible source-image synchronous dialogue/SFX, Seedance 2.0 for routine motion, and H3 for justified large movement. Use one operationKey per clip, require executionId, poll terminal state, and never re-create an unknown outcome.
  5. Prompt subjective action language, concrete scene/object anchors, body/camera motion, start/end state, sound behavior, and forbidden third-person framing. Avoid overloaded action and style-adjective piles.
  6. Inspect every accepted clip at first, middle, and last frames for POV integrity, action, references, screen direction, camera/body motion, duration, ratio, audio, and accidental protagonist reveal. Repair only the failed clip with a linked new key.
  7. Assemble accepted clips in storyboard order with combine-videos. Watch the completed film for continuous subjective grammar, narrative clarity, pacing, audio continuity, cut boundaries, technical integrity, and no unwanted watermark. Do not add a client-side publication confirmation; follow the API result.

User-facing output

  • Media type: One final MP4 plus POV spec, storyboard, continuity manifest, and operation receipts
  • Default quantity: 1 finished POV short
  • Content per video: A multi-clip first-person narrative with one continuous subjective contract
  • Default layout: User-requested duration, ratio, visual direction, and audio plan
  • Model policy: Kling Omni complex references; Kling 3 precise frames; Seedance 2.5 audiovisual/artistic; Veo eligible synchronous audio; Seedance 2.0 routine; H3 justified large motion
  • Downstream use: Human creative, continuity, safety, and publication review