Skip to content

Workflow Overview

Every RecapSynth project follows the same seven-step pipeline. This guide explains each stage in plain language—what it does, when you need it, and how it connects to the next step.

You do not have to complete all seven steps every time. For example, you might reopen a project and jump straight to Review after tweaking OCR in Write. For your first project, however, follow the order below.


RecapSynth sidebar navigation showing Import through Export
OrderNav LabelOne-Line Summary
1ImportLoad chapter images into the project
2MergeStitch or split pages for easier editing
3Cast

New

Detect character heads and cluster recurring characters
4WriteMark panels and write or prepare scripts
5ProcessGenerate AI scripts, audio, and crops
6ReviewCheck and fix generated content
7ExportSave an OTIO timeline for video editing

Nav label:Import·Detailed guides:Importing Chapters·Managing Chapter Pages

Import is where your raw material enters RecapSynth. You upload folders of chapter images—typically one folder per chapter, with each image file representing one page.

  • Upload chapter folders
  • View chapters in a list or grid
  • Reorder chapters with Auto sort
  • Open a chapter to manage individual pages (reorder, add, or delete)
  • Delete chapters you no longer need

Nothing else works without pages in the project—Import is always the first real step for a new recap. Before moving on, make sure every chapter you plan to recap appears in the chapter list with the correct number of pages.


Nav label:Merge·Detailed guides:Merging Pages·Splitting Long Pages

Comic and manhwa pages often arrive as many tiny images or a few extremely tall scroll strips. Merge gives you two modes—switched via tabs at the top—to reshape them:

Merge Mode

Combines selected pages into one continuous image (ideal for vertical scroll comics).

  • Select pages and click Merge Selected
  • Use Auto / Auto-Stitch to batch-process chapters
  • Switch between RAW and MERGED views

Split Mode (Slice)

Cuts overly tall images into smaller pieces that are comfortable to annotate.

  • Draw horizontal split lines on long pages
  • Use Auto-Detect to suggest split points
  • Adjust the Fetch Limit ratio to control which pages count as “long”

The Write , OCR , Head detections works best when pages are a manageable size. Skipping Merge is fine if your images are already one page per file at a reasonable height. Before moving on, confirm each chapter has pages you can comfortably view and annotate.


Cast is RecapSynth’s character-intelligence stage. It uses AI to find characters in your artwork, group recurring appearances together, and suggest names—so you don’t have to identify every face manually.

1. Head Detection

Scans pages and crops detected character faces for comparison.

2. Clustering

Groups visually similar faces into character clusters across chapters.

3. Name Extraction and Matching

Matches clusters to named characters from story text for confirmation.

4. Visual Assignment

Go through each page using an llm to match character name to speech bubble and describe what is happening in the art panel

  • Run Head Detection across the whole project
  • Run Clustering per chapter, then roll chapter clusters up into Characters (Super-Cluster) groups
  • Lock confirmed clusters so later runs don’t split them apart
  • Drag and drop clusters between character cards to assign or regroup them
  • Review AI-suggested names and approve or reject each suggestion

Accurate casting keeps character names consistent across your generated scripts and narration—especially valuable for long-running series with large casts. Before moving on, ensure each character cluster has a name attached to its cluster.


Write is the creative heart of RecapSynth. On each page, you mark what matters and pair it with a character assignment for each art and speech Ocr panels:

Speech Boxes

Dialogue and text bubbles read via OCR or manual entry.

Art Boxes

Visual action or scene panels that require narration and cropping.

  • Navigate chapters and pages
  • Run Detection to locate art panels automatically
  • Run OCR to extract text from speech boxes
  • Edit or draw rectangle or polygon boxes on the canvas
  • Manage Characters (names, gender, descriptions)
  • Edit script lines in the right-hand panel

Process relies on your boxes to know what to narrate—better annotations lead to cleaner scripts and fewer edits later. Before moving on, run Detection on the chapters you plan to process and verify scripts for key panels.


Nav label:Process·Detailed guides:Process Overview·Script Settings·Audio Settings·Crop Settings

Process runs AI and media pipelines on the chapters you select across three configurable tabs:

Script

Generates narration lines from your annotations using your configured language model.

Audio

Synthesizes voiceover audio through your configured text-to-speech provider.

Crop

Crops and enhances panel visuals (margins, contrast, and aspect ratio).

  • Select chapters to process
  • Configure settings across the Script, Audio, and Crop tabs
  • Click Start Processing and monitor the progress as the assets are created.

This stage produces the actual recap assets—narration text, voice files, and cropped panels—used in Review and Export. Wait for processing to finish and resolve any errors shown in the Task Processor console before moving on.


Nav label:Review·Detailed guides:Reviewing Scripts & Audio·Exporting Scripts·Uploading Audio

Review is your quality-control stage. You inspect every processed segment alongside its cropped image, script text, and generated audio.

  • Switch between chapter tabs
  • Click script lines to edit text inline
  • Play audio in the dock below the image preview
  • Export Script as TXT, Markdown, JSON, PDF, HTML, or CSV
  • Upload Audio to replace or add voiceover files in bulk
  • Clear all review data if you need to start over (destructive)

AI-generated scripts and voiceovers rarely come out perfect on the first pass. Listen to the audio, read scripts aloud, and fix any awkward phrasing before moving to Export (or export standalone scripts here if you aren’t building a video timeline yet).


Nav label:Export·Detailed guides:Export Overview·Motion Settings·Transition Settings

Export packages your processed segments, camera motion, and transitions into a single OTIO (OpenTimelineIO) timeline file compatible with professional video editors.

  • Set Project Name and Choose Destination
  • Choose which chapters to include
  • Configure Motion (camera movement on still panel images)
  • Configure Transitions (cross dissolves, slides, etc.)
  • Click Export Project, confirm, and wait for Export Complete
Setting Current Value
Format OTIO (fixed)
Frame Rate (FPS) 24.00 (fixed)

The Export Complete dialog lets you Show in folder, Open export folder, Go to Review, or Continue editing.


1Import — Upload all chapter folders
2Merge — Merge or split pages if needed (skip if page sizes are already manageable)
3Cast New — Run Head Detection, Clustering, and name assignment (optional, recommended for multi-chapter series)
4Write — Run Detection, draw missing boxes, assign characters, and run OCR on speech
5Process — Configure Script, Audio, and Crop settings, select chapters, and start processing
6Review — Refine scripts and listen to generated voiceovers
7Export — Set motion and transitions, then export your OTIO timeline

You can click any step in the left navigation rail at any time—RecapSynth does not lock steps behind each other.

Situation Suggested Jump
Need to add a missing chapter Import
Cast isn’t finished yet, but you want to start writing Cast → Write (re-run Cast on new chapters later)
Script needs edits, but boxes are fine Write (edit) → Process (re-run one chapter)
Audio is good, video pacing/effects need changes Export (adjust motion or transitions)
Starting over on generated assets Review → Clear, then Process again
Need to adjust character assignments after writing Cast (adjust clusters) (rerun Visual assigner)

Several AI and export steps consume credits, tracked live in the bottom-left credit chip:

Cast

Head Detection and Clustering display estimated credits before starting.

Process

Displays estimated max credits based on selected chapters and settings.

Export

Shows estimated credits directly in the export header.

Credits chip display showing estimated credit cost in the export header
The Credits Chip: shows estimated credits directly in the export header.

See Credits and Billing for full details on credit usage and plans.