Pipeline Overview
How WTN Suite's AI pipeline turns webtoon pages into narrated content
Pipeline Overview
WTN Suite works with two different versions of your artwork, and understanding the difference is the key to understanding the whole pipeline:
- Page — output of Webtoon Cropper. Still has text bubbles and dialogue intact. The AI reads pages, because it needs that dialogue for context when writing narration.
- Panel — a clean, text-free crop, produced automatically. This is what actually appears in your final video — the on-screen text is gone because the narration audio replaces it.
One manual step (Webtoon Cropper) turns your raw download into pages. From there, one click — Analyze & Narrate — runs three AI stages automatically: automatic panel cropping, scene analysis, and narrative writing.

How It Works
Your workflow:
| Step | What You Do |
|---|---|
| Step 1 | Run Webtoon Cropper on your downloaded scroll — produces pages |
| Step 2 | Click "Load Chapter & Review Panels" — loads your pages into the grid |
| Step 3 | Click "Analyze & Narrate" — AI crops clean panels, analyzes every page, and writes the narration |
What happens automatically when you click Analyze & Narrate:
| Stage | What It Does | Time Estimate |
|---|---|---|
| Panel Cropping | Crops each page down into a clean, text-free panel for video export | ~30 sec per page |
| Scene Analysis | AI reads each page — dialogue included — for characters, actions, mood | ~5 sec per page |
| Narrative Writing | AI writes a narration script for the entire chapter | ~30 sec total |
After the pipeline finishes, you export the narration to audio (using Edge-TTS or Kokoro-TTS), then create a video with the Video Compositor using the clean panels.
Before You Start
Make sure you have:
- Webtoon pages ready — download the chapter with WT-Downloader, then split the scroll into pages with Webtoon Cropper (required — the pipeline can't load a raw continuous scroll directly)
- An AI API key configured in Settings
💡 Tip
No character setup needed. WTN Suite v4.1.7 includes Auto Character Handling — the AI identifies characters automatically during scene analysis. No manual character building required to get started.
ℹ️ Info
How many images can I process at once? WTN Suite can technically handle 500–700 pages per run. However, we recommend processing one chapter at a time. Loading too many pages in a single run increases the risk of hitting your AI provider's rate limits mid-run, which wastes API requests and tokens without producing usable output.
Step 1: Crop Into Pages
Before anything else, run your downloaded chapter through Webtoon Cropper. This is the one manual step in the pipeline — WTN Suite needs pages, not a raw continuous scroll, to work with.
Place your cut lines at natural scene breaks. Where you cut matters: a clean cut gives the AI a self-contained page to read, while cutting mid-dialogue can split a character's speech across two pages and hurt narration quality later.
Step 2: Load Your Pages
Click "Load Chapter & Review Panels" on the toolbar and select the pages you just exported from Webtoon Cropper.
This step just loads your pages into the main grid — no AI runs yet. Each row shows:
- A preview of the page image
- Character list (initially "Awaiting Analysis")
- Narrative text (initially "Awaiting Narration")
- Status indicator
Step 3: Analyze & Narrate
Click "Analyze & Narrate" to kick off all three automatic AI stages.
Automatic Panel Cropping
WTN Suite crops each page down into a clean, text-free panel — this is what will show up in your final video. It handles this differently depending on your content:
| Mode | Best For | Set In |
|---|---|---|
| Manhwa Style | Korean webtoons (vertical scroll, clear panel borders) | Settings → Panel Detection Model |
| Manhua Style | Manga and manhua (page layouts with complex arrangements) | Settings → Panel Detection Model |

💡 Tip
If the automatic crop misses the mark or splits a page incorrectly, click on that row in the grid to open the Crop Editor. You can manually draw or adjust the panel boundaries by dragging the edges and corners.
Scene Analysis
The AI reads each page — dialogue and text bubbles included — and describes what's happening: characters present, their actions, emotions, the setting, and visual details. This analysis feeds into the Narrative Writing stage, and it's only possible because pages keep their text; a clean panel alone wouldn't give the AI enough to work with.

ℹ️ Info
Auto Character Handling (v4.1.7). The AI now automatically detects and tracks characters across pages — no manual character setup required. You'll see consistent character references throughout the narration. For even more precise control over character names, you can optionally use the Character Builder.
What the AI identifies in each page:
- Characters and their names (auto-detected)
- Actions and movements
- Emotions and expressions
- Scene setting and background
- Important visual details (text bubbles, effects)
Narrative Writing
Narrative Writing runs automatically after Scene Analysis finishes. The AI takes all the scene analysis results and writes a continuous narration script — a story that flows from page to page with smooth transitions.

Narrator Prompt Templates
You can choose how the AI writes the narration by selecting a template in Settings:
| Template | Style |
|---|---|
| Master Prompt v4.1 | Cinematic storytelling with continuity rules — works best for most content |
| Template 2 v4 | Choose between Pure Narration, Character Interaction, Visuals Only, or Mixed styles |
| Template 3 v3 | Simplified version of Template 2 — faster processing |
| Khmer Language | Narration in Cambodian/Khmer |
| Custom Prompt | Write your own instructions for full control |
💡 Tip
Start with Master Prompt v4.1 — it produces cinematic, engaging narration that works well for most webtoon genres. Experiment with other templates once you're comfortable with the pipeline.
Reviewing and Editing Results
After the pipeline finishes, review the results in the main grid:
- Edit narration: Click "Edit" next to any row's narrative text to modify it directly
- Re-run a single page: Click "Rerun AI" next to a specific page to regenerate its analysis and narration
- Export panels: Click "Export Panels" to save the clean, cropped panel images to a folder — this is what the Video Compositor uses
- Export narration: Click "Export Narration" to save all narrative text as a TXT file
💡 Tip
It's worth spending a few minutes reviewing the narration before generating audio. Small edits to the text (fixing character names, smoothing transitions) can significantly improve the final video quality.
Next Steps
Once you're happy with the narration:
- Generate audio — Use Edge-TTS (all plans) or Kokoro-TTS (Standard/Pro) to convert text to speech
- Create video — Use the Video Compositor to combine the clean panels + audio into a final video
- Advanced editing — Export to CapCut Director (Standard/Pro) for animations and subtitles
- Refine characters — Optionally use the Character Builder if you want to assign specific names to auto-detected characters