FAQ
Find quick answers to common questions about RecapSynth’s project files, Cast automation, script and audio generation, cropping, timeline exports, and troubleshooting.
General & Project Files
Section titled “General & Project Files”RecapSynth imports standard image files (JPG, PNG) for chapter pages and supports uploading external voiceovers in WAV, MP3, FLAC, or OGG (converted automatically to 16-bit PCM WAV). Video timelines export as OpenTimelineIO (.otioz) bundles, and script text can be exported to TXT, MD, JSON, PDF, HTML, or CSV.
All imported manhwa images, annotations, crops, and audio files are stored locally on your computer. When you export a video timeline, the Save Destination field on the Export screen controls where your .otioz bundle is saved.
Cast & Write
Section titled “Cast & Write”Cast automates character tracking before you reach Write: it detects faces inside panels, groups faces per chapter and across the whole manhwa, extracts character names from dialogue, and uses AI vision to tag which character spoke each bubble and write panel descriptions. Write is where you draw or refine dialogue and art boxes manually, run local OCR, and review or edit the tagged script lines.
Yes. You can go straight to Write, create characters in Character Manager, and tag each speech card manually using the character selector or shortcuts like Ctrl/Cmd + . and Ctrl/Cmd + /.
Use the Chapter Clustering (Chapters tab) or Character Grouping (Characters tab) screens inside Cast to merge face clusters within a chapter or combine chapter appearances across the entire series. Once visual groups are finalized, use Name Assignment to link names to each group.
Clicking it opens the Alias Evidence Visualizer. It jumps directly to the exact comic page and highlights the dialogue box where the AI discovered that name or alias, showing the chapter, page number, and mention type (1st Person, 3rd Person ID, 2nd Person, or Context).
In Cast → Script Assignment, a vision model looks at each page alongside your named character directory (plus Narrator and Unknown). It automatically matches each dialogue box to its speaker and writes descriptions for your art panels, which you can inspect and refine afterward in Write.
Process (Script, Audio & Crop)
Section titled “Process (Script, Audio & Crop)”- Flat Mode: Generates narration with strict 1:1 alignment to your detected panels and dialogue boxes in chronological order. Best for panel-by-panel recaps.
- Condensed Mode: Uses a two-pass AI workflow—first writing a smooth, continuous master script for the chapter (cutting filler and repetition), then mapping each sentence to 1–3 matching panels. Required if you want multi-panel layouts in Export.
By default, Process skips chapters that already have completed scripts, audio, or crops. Turning on an Override toggle (Override Previous Script, audio override, or Force Override in Crop) forces RecapSynth to delete existing outputs for the selected chapters and regenerate them from scratch. Note that overriding scripts also deletes linked voiceover audio for those chapters so your voice tracks stay aligned with the new text.
Yes. On the Process → Audio tab, choose Aether (Kokoro) or VoxCore (Supertonic-3) to synthesize dual-mono WAV narration directly on your machine—no external TTS API key needed. You can also choose Edge or cloud ElevenLabs, or skip AI TTS entirely and upload your own recorded audio in Review.
They style every panel crop in real time on the live preview stage: Canvas manages aspect ratio locking and blurred/solid background fills when panels don’t match your target video ratio; Crop controls expansion, cutouts, contrast boost, and speech text removal; Border adds rounded frames; and Shadow adds angled drop shadows for depth.
Export & Timeline Modes
Section titled “Export & Timeline Modes”- Flat: Places all panels sequentially onto a single video track (Video 1), one panel per cut.
- Layered: Places simultaneous panels onto multiple video tracks (Video 1, Video 2, etc.) using adaptive comic grids with uniform 32px gutters and coordinated camera motion.
- Auto (Default): Automatically exports as Layered if your chapters contain multi-panel scenes, or Flat if they do not.
Layered Mode only produces multi-panel track layouts when your script was generated in Condensed Mode (which maps 1–3 panels per sentence). If your script was generated in Flat Mode, each scene only has one panel, so Layered export outputs a single track just like Flat.
Troubleshooting
Section titled “Troubleshooting”- Make sure at least one chapter is selected in the chapter checklist.
- Check the Script and Audio tabs for a red Missing API Key alert and add your key under Settings → API Keys.
- Verify a valid model and voice are selected for every enabled stage.
Not all language models or providers support extended thinking budgets. If the Task Processor reports an unsupported reasoning parameter, lower Reasoning Level in Script → Advanced Settings to Low or None, and confirm your chapters have OCR/detection data.
Confirm Enable Audio is switched on in Process → Audio, verify the chapter has generated script lines in Review, and—if using cloud ElevenLabs—confirm your API key and ElevenLabs character quota are valid.
DaVinci Resolve opens .otioz bundles natively. If you use Premiere Pro or Final Cut Pro, confirm your OTIO import plugin is installed, or import into DaVinci Resolve first and export an XML timeline. Also ensure every exported chapter has synchronized voiceover audio in Review.