# Technical information

This describes the opening story film, Niji's Theater Episode 1, “Cheer Up, Dad!” Part 1. It is not a technical specification for the entire 23–25-minute tutorial.

| Item | Confirmed record |
|---|---|
| Japanese opening film | Approved v026, about 10 min 14 sec, 1920×1080, 24 fps, H.264/AAC. All 14,739 frames decoded. |
| English opening film | Approved v004, about 10 min 02 sec, 1920×1080, 24 fps. All 14,447 frames decoded. |
| Story assembly | Short scenes were previewed and revised with human decisions. Episodes were joined, overlaps removed, and boundaries checked. |
| Storyboard | The continuation draft explicitly requested a storyboard first. The working storyboard file is not published. |
| Images | Production records show OpenAI built-in image generation and image editing. The exact image-model ID was not recorded. |
| Video | Records for adopted shots show Google Gemini Interactions API, `gemini-omni-1.1-flash`, generating short clips from a starting image and text. We do not assert that every shot used the same model. |
| Narration and dialogue | Google Gemini API `gemini-2.5-pro-preview-tts`, voice `Sulafat`, appears in the adopted production records. Narration and dialogue were edited separately from generated video sound. |
| Sound and music | The ledger records a switch to the creator's own garden recording, `鈴虫.MOV`. The theme song has a Suno production record. Effects and ambience were edited separately. |
| Editing and captions | Python, NumPy, Pillow, and FFmpeg. Text was accurately composited later; local video and audio changes were versioned. |
| Quality checks | Full FFmpeg decode, boundary frames, audio and SHA-256 checks. faster-whisper assisted dialogue and pronunciation checks. Human viewing and listening determined acceptance. |

Japanese v026 and English v004 clarified Tako-san and Tako-tsuma in the bonus and postscript after the ending. They retained the preceding versions' audio while changing the specified displayed text. Interim Japanese v023 was rejected and is not a final specification.

AI planning or generation, deterministic editing/checking, and human judgment are separate tasks. Model and rights claims here are limited to the available records. Human review is still required before publishing franchise-inspired material.