# Sprite DX - Stage 3 - Splitting Shots

In this blog post, we explore ways to split 5 second video into multiple shots.

## Previously…

In [previous post](https://blog.sprited.app/designing-multi-shot-prompts-for-sprite-animation-generation), we designed prompt format for *Seedance 1.0 Pro* use in generating animation frames. We are using the following multi-shot prompt.

> \[Scene Description\] Character is a game sprite character for game called “Machi.” Her name is “Eliana.” The video has no camera movements.
> 
> \[SHOT 1\] Intro: Camera static full body shot and pixel art character says “hi.” \[CUT\]
> 
> \[SHOT 2\] Idle Loop: Camera static full body shot and the pixel art character is facing slightly right (+x) looking straight right (+x) and shows sprite animation loop for “idle” state, breathing in and breathing out, on a pure white background. \[CUT\]
> 
> \[SHOT 3\] Run Cycle (in-place, +x): Camera LOCKED on the pixel art character running in positive x direction, and character starts showing sprite animation loop for “run” state on a pure white background. \[CUT\]
> 
> \[Tags\] #character-animation #角色动画  
> #game-sprite #游戏精灵  
> #platformer #平台跳跃  
> #machi #Machi  
> #cute #可爱  
> #pixel-art #像素艺术  
> #1girl #少女  
> #subject-only #主体突出  
> #character-only #纯人物  
> #side-scroller #横版卷轴  
> #platformer #平台游戏  
> #idle-loop #待机循环  
> #run-cycle #跑步循环  
> #pure-white-background #纯白背景

It describes overall scene and then goes into 3 different shots—saying hi, idle loop, run loop. As a result, we get generations like the one you see here.

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1754073809953/63a53803-ab8e-40a4-adf7-0f6c9ed161ab.webp align="center")

## Problem

The problem is that this generated video clip now contains multiple shots (aka “cuts“). And to use it as a character animation sequence, we need to break the video down to multiple videos each containing one shot.

Let’s explore some options for splitting these shots into separate videos. From what I read from Seedance paper, they created a model that will split the shots. So, there may be open-source models that already do this.

---

### PySceneDetect

[PySceneDetect](https://github.com/Breakthrough/PySceneDetect) seems to be one of the most popular solutions here.

I vive coded a script to feed in the mp4 file and split into multiple shots.

We are using `threshold` value of `8.0`.

```bash
poetry run python ./scripts/loop_cut_video.py shots/shot-002.mp4 --out shot-002-loop.mp4
```

Source code: [https://github.com/kndlt/sprite-flux/blob/main/scripts/split\_shots.py](https://github.com/kndlt/sprite-flux/blob/main/scripts/split_shots.py)

**Shot 1: Saying Hi**

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1754074037423/82baff02-5a0b-4892-a9e0-5447a2d56740.webp align="center")

**Shot 2: Idle Loop**

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1754074098422/7dab1017-5eb3-407a-b657-2223e358a30e.webp align="center")

**Shot 3: Run Cycle**

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1754074120281/3b35bf37-dceb-4dce-9583-ceabae7d3d7c.webp align="center")

---

## Post Processing: Detect Loop

Splitting is working fine. Let’s now look at extracting animation cycles using the strategy explored in [previous post](https://blog.sprited.app/sprite-animation-generation-detecting-loops?source=more_series_bottom_blogs).

```bash
> poetry run python ./scripts/loop_cut_video.py shots/shot-002.webp --out shot-002-loop.webp

Loop: 18->30 len=12 seam=0, avg_mot=746.245, score=0.001
Saved shot-002-loop.webp
start:      18
end:        30
length:     12
seam:       0
avg_motion: 746.2448
score:      0.0006700213998556137
fps:        23.80952380952381
```

Source code: [https://github.com/kndlt/sprite-flux/blob/main/scripts/loop\_cut\_video.py](https://github.com/kndlt/sprite-flux/blob/main/scripts/loop_cut_video.py)

We will ignore “saying hi” shot since that one is just a dummy one.

## Final Outcome

**Idle Loop**

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1754075071098/56e04399-916f-400f-925d-1f0042b2f840.webp align="center")

**Run Loop**

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1754075003820/e7afad04-c8c2-48f9-99f7-9ad2927e23c0.webp align="center")

---

## Conclusion

We now have series of python scripts to run to extract the animation clips. In next post, we will build custom nodes for ComfyUI.