How to work with subtitles: local transcription and SRT export
The 'Subtitles' tool recognizes audio directly in the browser and outputs an SRT file with timestamps. This is convenient for quick transcription of short recordings, but the first result almost always requires manual editing. In this guide we will explain how to get a draft, check synchronization, fix text and save the ready file. We will not attribute functions to the tool that it does not have: there are no settings panels, batch processing, translation or automatic saving. Everything you do is upload audio, press 'Generate' and work with one result in a separate panel. Examples in the article are fictional and provided only for illustration.
Open tool: Subtitle generator ↗What the tool actually outputs and how to work with it
After pressing the 'Generate' button you get one result: text with timestamps in SRT format. This file can be downloaded or copied. Recognition takes place locally in the browser, audio is not sent to external servers. This means that speed and quality depend on your device and recording length. Limits: audio up to 20 MB and up to 5 minutes, text up to 3000 characters for speech. If the recording is longer, it must be split into parts and processed separately, then SRT files manually merged.
The first result is a draft. It may contain errors: wrong words, omissions, merging or splitting of phrases. Your task is not to regenerate everything anew, but to make targeted edits. The tool has no built-in subtitle editor, so editing is done in an external text editor. Download SRT, open it in Notepad or a specialized program and work with the content. This way you will preserve timestamps and structure.
Do not try to upload audio again if the error is local. It is better to fix one file than to run the system. Remember: the model can make mistakes, and queue and availability vary. If the result does not appear, perhaps the limit has been exceeded or the service is temporarily unavailable. In this case reduce the file size or split it into fragments.
Example: you uploaded a 4-minute lecture recording. The tool output SRT with 30 replicas. You downloaded the file and opened it in an editor.
Checking timestamps and synchronization
In SRT each subtitle has a number, start and end time, and then text. The first thing to check is whether the time matches the actual sound. Play the audio and SRT in a player that supports subtitles. If a subtitle appears too early or too late, fix the time manually. Time format: hours:minutes:seconds,milliseconds. For example, 00:01:23,456. Do not change the format, otherwise the file will not be recognized.
A common problem is long pauses inside one subtitle. The tool may merge two phrases into one if there was no clear silence between them. Split such a subtitle into two, adding a new timestamp. Conversely, if a subtitle is too short and cuts off mid-word, check if the ending was omitted. It may be worth merging it with the next one.
If you split the audio into parts, after merging SRT files check the junctions. Timestamps may get messed up if you did not account for the shift. For example, the second part starts at 00:00:00, but it should continue from 00:05:00. Fix this manually, otherwise subtitles will be out of sync.
Example: in SRT subtitle 5 starts at 00:02:10,000, but in the audio the phrase sounds at 00:02:12,500. You change the start time to 00:02:12,500 and the end to 00:02:15,000.
Text editing: spelling, punctuation, names
Local transcription is not perfect: it may confuse similar words, especially proper names, terms and abbreviations. Review the entire text and correct errors. If the audio had clear pronunciation but the tool made a mistake, replace the word. Do not rely on auto-replace: it may ruin timestamps if you edit in a program with automatic formatting. Better to use a simple text editor.
Punctuation in SRT is not mandatory but improves readability. Add commas, periods, question marks. If speech is fast, split long sentences into shorter subtitles. Remember: subtitles should be easy to read from a screen. Optimal line length is 35–42 characters. If a line is longer, move part of the text to a new line within the same subtitle. In SRT a line break is done by simply pressing Enter inside the text block.
If there are multiple speakers in the recording, you can add their names at the beginning of a subtitle, e.g. 'Anna:'. However the tool does not separate voices, so determining the speaker must be done manually. This is laborious but necessary for interviews and dialogues. If separation is not required, leave the text as is.
Example: the tool recognized 'privet kak dela' as 'privet kak dela'. You add a comma: 'Privet, kak dela?' and split into two lines: 'Privet,' and 'kak dela?'.
Working with language and localization
The tool allows you to select the output language. If you select the wrong language, the result will be meaningless. Check that the language matches the audio. If the recording contains mixed languages, recognition may work worse. In this case process fragments separately, then merge SRT. Do not try to translate subtitles within the tool: it is not designed for translation. Use separate services for translation, but remember that timestamps must be preserved manually.
If you want subtitles in another language, first get the text in the original language, then translate it and insert the translation into SRT, preserving timestamps. This is double work, but there is no other way. The tool does not perform translation and does not have a subtitle duplication function.
When working with Cyrillic or other non-Latin alphabets make sure the file encoding is UTF-8. Otherwise characters may display incorrectly. Most modern editors support UTF-8 by default, but check before saving.
Example: audio in Russian, but you accidentally selected English. Result – a set of meaningless words. You re-generate with Russian, then correct text.
Final check and export
After editing save the file with .srt extension. Make sure the structure is not broken: each block separated by a blank line, subtitle numbers in order, time in correct format. Open the file in a player and view several fragments. If subtitles display and are synchronized, work is done.
Do not forget that the tool does not save history. If you close the tab, the result disappears. Download SRT immediately after generation. If you need to make several edits, save intermediate versions under different names so you can revert.
Remember the limits: maximum audio size 20 MB and 5 minutes. If the recording is longer, split it. Also keep in mind that the model can make mistakes, and queue and availability vary. Do not rely on the tool for critical tasks without human verification.
Example: you saved SRT, opened it in VLC and made sure subtitles appear on time. Done.
Typical errors and how to avoid them
Error 1: regenerating instead of editing. If the result is basically correct, do not waste time on re-running. Edit manually. Error 2: ignoring timestamps. Even perfect text is useless if it does not appear at the right time. Error 3: using complex editors that change SRT format. Use simple text editors.
Error 4: uploading audio exceeding limits. The tool will not process audio larger than 20 MB or longer than 5 minutes. Split the recording in advance. Error 5: expecting translation or voice separation. These functions are not available. Error 6: trying to load a document for spelling check. The tool does not work with documents, only with audio.
If you encounter a server error, wait and try again. Perhaps the queue is full. Do not send the same file repeatedly – this will not speed up the process.
Example: you uploaded a 25 MB file. The tool returned an error. You split it into two parts of 12 MB and processed separately.
Editing subtitles after local transcription is manual but necessary work. The 'Subtitles' tool gives a draft, and you turn it into a ready SRT. Check timestamps, fix text, watch encoding and remember limits. Do not expect the tool to have functions it lacks: it does not translate, does not separate voices and does not save history. However it works in the browser, does not send audio to a server and outputs result in a convenient format. Use it as a first step, and final quality depends on your attentiveness.
Open tool: Subtitle generator ↗Interface translations are machine-assisted and may need refinement.