Workflow steps
Each step below is one card in a workflow. Cleaning, translating and syncing keep each subtitle's identity and timing; the clean-up rules may drop, merge or re-time cues exactly as they do on the Clean Up screen.
Inputs
| Video or audio files | Files and folders of .mp4 .mkv .avi .mov .wmv .flv .webm .m4v .mpg .mpeg, or sound-only .mp3 .wav .m4a .aac .ogg .flac .wma. Audio goes to Transcribe, Extract audio and Inspect; Embed and subtitle extraction need a video. |
|---|---|
| Subtitle files | .srt .vtt .webvtt .ass .ssa, read with their italics, colours and positions and their lines as written. |
| Image subtitle files | PGS .sup, VobSub .idx with its .sub (naming either, or both, is one file) and MicroDVD .sub. |
Media steps
| Inspect media | Reads audio, video and subtitle tracks. Needs FFmpeg. |
|---|---|
| Select audio track | Picks one audio track per file: the default (never a commentary), by language, or by number — with fail, use the default or use the first when a file has no such track. The reason is shown and a fallback is logged. |
| Extract audio | Writes the chosen track as WAV, FLAC, MP3 or AAC (M4A), with optional sample rate, mono or stereo and bitrate. Needs FFmpeg. |
| Extract subtitles | Pulls a text subtitle track out of a video (first, by language or by number; a full track beats a forced or SDH one). Needs FFmpeg. |
| Extract image subtitles | Pulls a PGS or VobSub track out as its image file, to feed Convert to text. Needs FFmpeg. |
| Embed in video | Adds one or two connected subtitle files — and, if asked, the subtitle files beside the video — as tracks of a new copy beside the video, without re-encoding. Needs FFmpeg. |
Subtitle steps
| Transcribe | Speech recognition on this computer with the language, installed model and vocabulary hint you set. Timing and line layout follow Settings. Honours the chosen audio track. Needs FFmpeg, a speech engine or Whisper with a model, and a license. |
|---|---|
| Convert to text | The Convert screen’s conversion: OCR for PGS and VobSub, plain parsing for MicroDVD. One file at a time. Needs a license. |
| Clean up | The Clean Up tool’s rules (tags, credits, SDH, overlaps) and presets. No AI. |
| AI clean up | Fix errors, improve flow or tighten, with an optional context note. Uses your own DeepSeek key; credits cannot pay for it. Needs a license. |
| Shift timing, Shift a section | Move every subtitle, or a range counted from 1, earlier or later. |
| Colour, Position | The Tools screens’ restyling, for all subtitles. |
| Translate | DeepSeek with your key or with credits — whichever Pay with says. From and to language, style, context note and metric conversion. Transcripts are corrected while translating. Timing and order are kept. |
| Synchronise | ffsubsync against a video’s audio (the chosen track), using the connected video or the one beside the subtitle file. Needs FFmpeg, ffsubsync and a license. |
Output steps
| Create subtitle file | Writes the subtitles the same as the input (the default: an ASS file keeps its styles and signs, a transcript is SRT), or as SRT, WebVTT, ASS or text; optionally re-wraps lines to the Settings layout. |
|---|---|
| Export | Saves to a folder or next to the original, with a file-name template and an explicit rule for existing files. This is the only step that writes your files. |
Clean up, Translate and AI clean up work on plain text, as their screens do, because a model reads <i> as words. Shift, Colour and Position keep the formatting.
FAQ
Why is a step orange?
Something it needs is missing or wrong — a setting, a connection, an engine, or a license. The strip under the row says what.
Why does Transcribe say "Also uses step 1"?
It takes its video from the input step and its audio track from Select audio track, so it reads from more than the card on its left.
Can I drag steps from the shelf?
No. Steps are added by clicking, because the window's file-drop handling blocks dragging inside the page on Windows.