# Technical information

## Finished story

| Item | Value |
|---|---|
| Duration | 75.00 seconds |
| Shots | 25 shots, 3 seconds each |
| Video | 1280×720, 24 fps, H.264 |
| Audio | AAC, 48 kHz, stereo |
| Video review | 9 frames per shot, 225 frames total |

## Models and tools used

| Purpose | Model or tool | Role |
|---|---|---|
| Story structure | `gemini-3.6-flash` | Organized the source into shots |
| Storyboard images | `gemini-3.1-flash-image` | Previewed framing, backgrounds, and character placement |
| Character corrections | OpenAI image generation | Corrected faces, hats, acting, and unwanted characters |
| Three-second clips | `gemini-omni-flash-preview` | Generated motion from opening frames and character references |
| Speech | `gemini-3.1-flash-tts-preview` | Generated dialogue and narration by speaker |
| Speech verification | `gemini-3-flash-preview` | Checked missing or outdated lines |
| Video and audio | FFmpeg / `imageio-ffmpeg` | Normalization, concatenation, mixing, and subtitle burn-in |
| Exact text and cards | Python / Pillow | Rendered exact Japanese text |
| Storyboard workbook | `openpyxl` | Built the Excel review workbook |
| Change verification | SHA-256 | Detected unintended changes to approved material |

Model IDs describe the production record at that time. A new production must check the current validated defaults in Video Studio and current provider information.

## Quality decisions

- Final speech was produced separately from generated video audio.
- Exact Japanese text was rendered in post-production.
- Every shot was normalized to three seconds and 72 frames before concatenation.
- Character reference images were supplied for shots containing registered characters.
- Nine points inside every clip were checked for changes in faces, hats, text, and expressions.
- The last date correction changed one shot while preserving the other 24 shots and the approved audio.