> **⚠️ SUPERSEDED** — This document is a historical receipt. See `s7-current-understanding.md` for the authoritative current position.

# Totally Spies S7 — executive summary

## What we set out to answer

We wanted to determine whether Season 7's animation style is simple enough that
an image-first AI workflow would be sufficient, or whether a custom
video-oriented pipeline is materially necessary.

## What we did

We built and reviewed a local official reference set from Season 7, including:

- trailers
- clips
- compilations
- full episodes
- character promos

We downloaded **104 official videos**, extracted catalog stills and motion
snippets, and tested the conclusions against multiple local multimodal models.

## What we found

### 1. Season 7 is not full hand-drawn fluid animation

The show consistently reads as:

- clean, digital 2D linework
- flat color fills
- digital cutout / rigged animation
- controlled, restrained motion

### 2. The challenge is restraint, not complexity

The main AI problem is not generating spectacular realistic motion.
It is preserving:

- stable faces
- stable costumes
- stable linework
- held frames that remain still
- readable gadgets during action

In other words, the system needs to learn the franchise's **motion discipline**,
not just its visual design.

### 3. A still-image pipeline is not enough

A generic frame-by-frame image workflow is likely to fail by introducing:

- flicker
- face drift
- costume drift
- background drift
- unwanted extra motion

That would make the result feel like generic AI animation rather than Totally
Spies S7.

### 4. A custom video pipeline is justified

The business case is therefore not:

> “AI can make motion.”

It is:

> “Cultshot can make motion that stays on-brand over time.”

That is the distinction that matters for Banijay.

## What matters most for validation

We now have three benchmark packs to evaluate any future generation system:

1. **Character / costume consistency**
2. **Dialogue / hold stability**
3. **Action / gadget control**

Any system that fails identity stability or held-frame stability should be
rejected before deeper investment.

## Strategic conclusion

If the goal is to produce reviewable Totally Spies-style marketing video,
then the highest-value investment is a pipeline optimized for:

- temporal consistency
- identity persistence
- restrained cutout-style motion
- structural stability in action beats

not a pipeline optimized mainly for generic video realism.

## Bottom line

The evidence supports continuing with a **video-native, temporally controlled
Cultshot workflow** rather than an image-only approach.