How to edit video subtitles locally — SRT/ASS, not translation

The short answer

To edit video subtitles locally, import an SRT or ASS file or type the cues yourself, adjust each cue's in and out points, style, and position on a live preview, then burn them into the picture or export MKV soft subtitles. It edits subtitles you bring; it does not transcribe speech or translate. Use Subtitle Edit, with no upload.

What the editor does

• Add cues by hand or import an SRT/ASS you already have
• Adjust in/out, style, and on-picture position with a live preview
• Burn captions into the picture, or export MKV soft-subs for a client who wants a toggleable track

What is not shipped

• Translation — including any “localize” path that auto-translates the language track
• ASR / speech-to-subtitle — KwaFlux does not transcribe audio into cues
• Subtitle effects or a caption SaaS — this is an editor, not a motion-graphics pack

How to edit the cues

1. Open Subtitle Edit and import the clip you will caption
2. Type cues or import SRT/ASS, then style and position on the live preview
3. Check a 1s/3s/5s preview on the densest line, then export burn-in or MKV soft-sub locally

Cues you bring, not a transcript

Subtitle Edit is a cue editor on your GPU and disk. You import an SRT or ASS you already have, or you type the lines. You then set in and out, style, and on-picture position against a live preview. The delivery is either a burn-in, where the words are drawn into the picture, or an MKV soft-sub, where the client can toggle a track. That is the whole job in this release.

It does not translate. There is no path that reads one language track and writes another, including anything described as localizing the dialogue for you. If the cue file is in the wrong language, fix it in a translation step you do outside this module, then import the corrected SRT or ASS. The editor will style and time what you give it. It will not choose the wording.

It also does not turn the soundtrack into cues. Speech-to-subtitle is not in the current release. A clear interview, a noisy street, or a music video all have the same limit here: if you do not already have a subtitle file, you type the lines while you watch. Audio denoise can clean a soundtrack in its own module. That cleanup still does not produce captions. Do not expect the two modules to hand text to each other.

Timing and placement are where the preview earns its keep. Park the free 1s, 3s, or 5s window on the densest line and on a cut where the speaker changes. A line that fits a wide frame can collide with a face, a lower third, or the edge after a reframe. Burn-in is the right export when the destination flattens tracks or the client must not be able to turn the words off. MKV soft-subs are the right export when they want a toggle. Check both the in-point and the out-point. A cue that lingers into the next shot is a timing error, not a style error.

The screenshot on this page shows three cues typed in by hand, with the current one over the preview. Nothing has been exported yet, so it is not a burn-in. Import your clip, bring or type the cues, and judge the timed preview on your own line. The preview needs no account. Sign-in and a paid plan are needed to export, convert, repair a file, or download a file from a link. Processing runs on your machine.

Decide the handoff before you spend time on style. A burn-in cannot be undone by the client, so confirm spelling, line breaks, and the safe area on the densest cue. A soft-sub can be swapped later, which is the handoff to choose when the wording might still change. Either way, the words have to exist first. KwaFlux will not invent them from the audio, and it will not translate a cue you imported in another language. If you are also repairing the picture or the soundtrack, do that in the matching module and bring the finished media here. Subtitle Edit does not upscale, denoise, or replace a background. It places text you already wrote.

• Import SRT or ASS, or type cues. The editor does not translate them.
• Speech-to-subtitle is not in this release. The soundtrack is not a cue source.
• Burn-in draws words into the picture. MKV soft-subs stay a toggleable track.
• Preview the densest line and a speaker change before you export.

Further reading: Short-drama localization on one GPU: most middle-processing steps do not need a cloud pipeline.