DANIEL'S CREATOR TOOLS / 24 SEPTEMBER 2026
Caption Desk: from words to a finished subtitle file
Caption Desk is a working subtitle editor. Import an existing SRT or basic WebVTT file, correct it while watching your local video, and export a full subtitle file or captions timed for a shorter clip. You can work alone. There is no course to complete and no group to assemble.
Your first five minutes
- Open the desk. The fictional 66-second example is already loaded.
- Select caption 2. Its start is 00:00:04.100, but caption 1 ends at 00:00:04.200. Change caption 2’s start to 00:00:04.200 and press Save caption. The overlap count falls to zero.
- Select caption 4. Its half-second duration is flagged. This deliberately demonstrates a review prompt; without actual audio, there is no defensible automatic timing correction.
- Under Prepare one clip, use the supplied 00:00:12.400 to 00:00:30.000 range. Review it, then download clip SRT. Four captions cover a 17.6-second cut, and the first begins at zero.
- Download a backup. It preserves your editable captions and project name.
Bring in a real recording
Obtain subtitles from your existing transcription or video workflow. Import a UTF-8 SRT or VTT file smaller than 2 MB with at most 2,000 cues. Attach the corresponding video with Open local video. The browser reads it from your device; the application does not upload it. Playback depends on browser codec support.
Use a subtitle file timed against exactly the same video version. A trimmed intro or different export speed can make all captions appear wrong. A constant mismatch can be corrected with an offset; drift that increases over time needs timing work in the original editing/transcription workflow.
The controls, plainly
| Control | What it actually does |
|---|---|
| Save caption | Saves edited text and start/end times, then reorders captions by start time. |
| Wrap lines | Breaks text at word boundaries using your character limit. It never truncates words. Long words can still exceed the limit. |
| Split in two | Divides text near its word midpoint and estimates the timing in proportion to word count. Review the result by ear. |
| Merge next | Joins the selected caption and its neighbour, including any intervening gap. |
| Offset | Moves every caption by the same number of seconds. A negative start is rejected. |
| Undo | Restores up to 30 recent caption/project changes in the current session. It does not persist after reload. |
| Full SRT / VTT | Downloads plain captions with original source timing. It does not export video or visual styles. |
| Clip SRT | Keeps captions intersecting the chosen source range, trims their boundary times, and subtracts the source start from every timestamp. |
A practical church-media workflow
Open the subtitle file for one sermon recording. Check the speaker’s names, Bible references, negatives such as “not,” and any Telugu/English switches against the audio. Do not convert an uncertain transcription into a confident quotation. This tool does not verify Scripture or translate languages.
Listen before and after a promising clip. Decide whether the excerpt retains the question, qualification and conclusion. Set the exact cut range in Caption Desk and your editor. If the range cuts through a spoken word, adjust it to a natural boundary. Export clip SRT and import that file alongside the cut video in software that accepts basic SRT. The programme used to cut the video is responsible for font, colour, placement and burned-in captions.
For a training demo, follow the same process: verify object names, button labels, numbers and action order. A visually tidy caption can still contain an incorrect instruction.
What a review flag means
Default prompts are 20 displayed grapheme clusters per second, 42 per line, no more than two lines, and a minimum cue duration of 0.8 seconds. These are authored starting settings for manual review, not universal accessibility standards, YouTube rules or accuracy scores. You can change speed and line length. Indic scripts use grapheme segmentation when the browser supports it; otherwise counting falls back to Unicode code points.
Overlaps can be intentional when different people speak. Caption Desk flags them for review; it does not silently erase them. Empty captions and arrow timing markers inside text block exports. Invalid timing blocks imports. No automatic tool can establish whether the text matches what was said.
Format and privacy boundaries
- Basic SRT and VTT text/timing are supported. Styling, VTT cue positioning, voice tags, notes, regions and style blocks are omitted with an import notice. Preserve the original if you need those features.
- A project backup contains caption text and times, not the video. Reattach the video after reloading or switching devices.
- The current project is saved in this browser when storage is available. There is no cloud sync. Clearing site data, using private browsing or changing browsers may remove access to that copy. Download JSON backups.
- The hosted desk is private. Application code sends no subtitle or video content to a server. Hosting still handles ordinary page requests. The offline HTML in the accompanying package makes no network requests.
- Preview shape changes the preview box only. The app does not transcribe audio, translate, change colour, crop a frame, render a clip, remove a watermark or publish to any platform.
- Mobile browsers support text editing; a desktop is easier for video review and handling downloaded SRT/JSON files. On Android, open the private site in your browser and use the Downloads app to locate exports.
Sources & implementation choices
Reviewed 24 September 2026. W3C WebVTT specification informed cue timestamps, timing and plain-text export. It describes timed text and richer features beyond the intentionally basic support here. YouTube supported subtitle files confirms support for basic SRT and UTF-8 text. Compatibility with every editor/version is not claimed. Import the exported file and inspect it in your actual workflow before publishing.
The practice transcript is original fictional editorial text. It is not a real sermon, a Bible quotation or a transcription of Daniel. The tool has no affiliation with W3C, YouTube or Filmora.
Return to the working desk