How to Use Auto Subtitle Generator Offline: Step-by-Step
Reviewed by the OnlineFree.app team · Updated
Key points
- Auto Subtitle Generator Offline times a transcript you already own; it does not transcribe audio.
- Everything runs locally in the browser tab, so no media or text is uploaded.
- Files must be 100 MB or smaller and in mp3, wav, m4a, mp4 or webm format.
- Min gap (350 ms default), max line length (42 chars) and offset (±5000 ms) control the output.
- Diagnostics show aligned versus unaligned lines, revealing transcript or audio mismatches.
What Auto Subtitle Generator Offline actually does
It is a browser-only subtitle timing studio: you drop in a video or audio file plus a transcript you already have, it detects speech segments locally, aligns your words to those segments, and hands back a downloadable .srt or .vtt. Nothing is uploaded to a server, and the tool deliberately does not do speech-to-text — the words still come from you.
That split is the whole point. Most captioning tools do two jobs at once: guessing the words, and timing them. Guessing is the expensive, error-prone half; timing is arithmetic once you know where speech starts and stops. If you already own the words — a recorded script, a YouTube auto-caption export, meeting notes — you can skip the first job and only need the second. Auto Subtitle Generator Offline is built around that narrower task.
Because the media never leaves your tab, it suits client footage, unreleased course modules and anything under an NDA. The trade-off is that your browser has to decode the file itself, so keep it at or under 100 MB and in a format the browser can read: mp3, wav, m4a, mp4 or webm.
How do you use it, step by step?
Step 1: get your transcript into a pasteable form. One caption line per line of text works best, but you do not have to be precise — the tool will split on newlines first, then on sentence punctuation. If you paste fewer than five words, it warns you before wasting your time.
Step 2: drop the media file onto the left panel. Use the smallest file that contains the real audio; a 100 MB cap means a 40-minute 1080p export often will not fit, while the extracted mp3 of the same video usually comes in well under 20 MB.
Step 3: paste the transcript and choose a timing mode — segment per line, per sentence, or per paragraph. Step 4: watch the waveform on the right, where detected speech blocks sit against your cue spans; clicking a cue seeks the player so you can hear that exact in-point.
Step 5: nudge the global offset if everything is uniformly early or late, then click the SRT or VTT button. The download is named from your source file, so a clip called interview-final.mp4 produces interview-final.srt. The same no-upload, no-account approach runs across the other free online tools on OnlineFree.app — including a free invoice maker if you need one after the edit is delivered.
Tuning the timing controls
Min gap defaults to 350 ms and is the silence length used to decide where one speech segment ends and the next begins; it accepts 100 to 2000 ms. Lower it toward 150–200 ms for fast, clipped delivery where sentences run together. Raise it to 500–700 ms for interview audio with breaths, lip smacks and room tone, which otherwise register as separate speech blocks and fragment your cues.
Max chars per line defaults to 42 characters, adjustable from 20 to 80. That default keeps a cue to roughly two readable lines on a phone. If your target is a vertical video with a large caption font, drop it to 30–35. If you are timing a documentary for a wide screen, 50 is comfortable. The wrapping happens after alignment, so changing it never disturbs your in/out timecodes.
Offset in ms shifts every cue at once, from −5000 to +5000. Use it when the whole file is consistently ahead or behind — for example, if the recording has a two-second lead-in before anyone speaks, a +2000 ms offset moves every cue later in one step rather than by hand.
Timing mode decides how your pasted text is chopped before alignment: per line for pre-broken caption files, per sentence for prose transcripts, per paragraph when you only need rough chapter-level timing.
What the diagnostics strip tells you
Below the cue table, the diagnostics strip reports three numbers: detected speech segments, aligned versus unaligned transcript lines, and whether any cue implies an uncomfortably fast reading pace. The middle number is the one to act on.
If you paste 400 words but the detector finds only a dozen speech segments, you have far more text than audio. Unaligned lines are usually the overflow — extra sentences with nowhere to sit. Either trim the transcript to what is actually spoken, or switch timing mode so more words share each cue.
If the segment count is much higher than your line count, background music, keyboard noise or room hum is probably being read as speech. Raising min gap to 600–800 ms is the usual fix. The reading-pace warning flags cues that pack too many words into too few seconds; splitting those lines by hand is faster than re-running the whole alignment.
One habit worth keeping: play the first cue, the last cue, and any cue where two speakers overlap. Those are where timing errors hide, and no automatic tool we have used catches all three.
Limits, mistakes and when to verify
The honest limitations first. This tool does not transcribe — bring your own words or use a speech-to-text service first. The file must be 100 MB or smaller, and as of 2026 mkv containers with AC3 audio and DRM-protected streams typically fail to decode in a browser, so remux to mp4 or extract the audio to mp3.
The tab must stay open until the download finishes; closing it mid-alignment loses the work. Long files also take longer to decode, since the work happens on your machine rather than on a server.
The output formats are plain, well-documented text. SRT follows the long-standing SubRip structure that YouTube and most editors accept, while VTT is the W3C's WebVTT format and adds a header line plus optional styling hooks.
Finally, treat the export as a strong first pass, not a guarantee. If subtitles are going to air, to a paying client, or to a platform with accessibility obligations, spot-check the timing in your player of choice and adjust the offset before publishing.
Frequently asked questions
Does Auto Subtitle Generator Offline convert speech to text?
No. It does not perform speech recognition at all. You supply the transcript yourself — from a recorded script, a YouTube auto-caption export or meeting notes — and the tool handles only the timing, matching your existing words to the detected speech segments and exporting a timed .srt or .vtt file.
Is my video or audio uploaded anywhere when I generate subtitles?
No. Auto Subtitle Generator Offline runs entirely inside your browser tab, so the media file and transcript never leave your device. That makes it usable for client footage, unreleased content and anything covered by a confidentiality agreement, provided your browser can decode the file format.
What file formats and file size does Auto Subtitle Generator Offline accept?
It accepts audio and video the browser can decode — mp3, wav, m4a, mp4 and webm — with a maximum size of 100 MB. As of 2026, mkv containers with AC3 audio and DRM-protected files usually fail, so remux or extract the audio to mp3 before uploading.
How do I fix subtitles that drift out of sync?
If every cue is consistently early or late, use the offset field, which shifts all cues together and accepts values from −5000 to +5000 ms. If only some cues drift, the issue is usually segmentation: raise or lower the min gap setting, which defaults to 350 ms, and re-align.
Is the exported .srt accurate enough for YouTube or a course platform?
It is accurate enough for a first pass, and normally needs no more than a spot-check. Play the first cue, the last cue and any overlapping dialogue, then adjust the offset if needed. For broadcast or paid client work, always verify the timing in your final player before publishing.