Transcription vs translation vs dubbing, SRT vs VTT vs burned-in, a workflow that stops errors multiplying, Netflix line-length rules and Hebrew RTL.
A 40-minute lecture in Hebrew, with good captions in English, Spanish and Russian, reaches people the Hebrew version never will. Getting there is not one step. It is three: turn the speech into accurate text, turn that text into other languages, and fit the result into lines a viewer can read in the time they are on screen. Bad translated captions often fail at the first or the last step, not at the translation itself.
This guide walks through the difference between transcription, translation and dubbing, which caption format to use where, a workflow that keeps errors from multiplying across languages, the line-length and reading-speed rules professional subtitlers follow, what changes for Hebrew and other right-to-left languages, and how to think about cost and time.
These three get mixed up, and tools are often sold under the wrong name.
| What it produces | What the viewer gets | Typical use | |
|---|---|---|---|
| Transcription | Text of what was said, in the same language, with timing | Captions in the original language | Accessibility, muted viewing, search, and the source for everything else |
| Translation (subtitles) | The transcript rendered in another language, with the same timing | Captions in their language; the original voice stays | Reaching other-language audiences while keeping the speaker's voice |
| Dubbing | A new audio track in another language, recorded or synthesised | They hear their language | Children's content, entertainment, audiences that do not read subtitles |
Translated subtitles are cheaper and faster than dubbing, and they keep the speaker's own voice, which matters for a teacher, a rabbi or an interviewee. Dubbing is a different product with different problems (voice consent, lip sync). This guide is about subtitles.
A caption is either a separate text file that the player shows on top of the video, or text drawn into the picture itself.
| Format | What it is | Use it for | Limits |
|---|---|---|---|
| SRT (SubRip) | Plain text: a number, a start and end time, the text | YouTube, Vimeo, most editors and players; the most widely accepted | No styling or positioning, per YouTube's own format list |
| VTT (WebVTT) | Similar text format made for web players | Web players, HTML5 video, YouTube | Supports positioning, limited styling |
| Burned in | Text rendered into the video frames | Shorts, Reels, TikTok, WhatsApp, anywhere a caption file cannot be attached | Cannot be turned off or switched to another language; one file per language |
YouTube accepts uploaded caption files (SRT and WebVTT are both on its list of supported formats) and lets you add one track per language: in YouTube Studio, open the video under Subtitles and use Add language for each one. A viewer then picks the language in the player. Note what YouTube's help page says about its own automatic captions: they are in the video's default language only. So for a Hebrew video, uploading your own English, French and Russian files is the way to get those languages as proper tracks. Sources: Add subtitles and captions and Supported subtitle and closed caption files.
Burned-in captions are the right choice for vertical clips, because the feeds of Shorts, Reels and TikTok play the video as it is. The cost is one export per language.
A practical rule: long video on YouTube or your website, upload files, one per language. Short vertical clips, burn in, usually in the language of the audience you post to.
Every error in the source transcript is copied into every translation. A misheard surname in Hebrew becomes a wrong surname in eight languages, and you will not be able to spot it in Russian. So the order matters.
Automatic speech recognition is good on clear speech from one microphone and gets worse with crosstalk, room echo, music under speech and specialist vocabulary. Give it the best audio you have (the direct recording, not a re-recorded copy from a phone).
Read the whole transcript against the video once, before any translation. Look for:
Before translating, write down every name and term and how it should appear in each target language. A short table is enough:
| Source | English | Spanish | Notes |
|---|---|---|---|
| (the speaker's name in Hebrew) | The agreed Latin spelling | Same as English | Use the spelling the person uses themselves |
| Gemara | Gemara | Guemará | Do not translate as "Talmud" unless you mean it |
| (a project or product name) | Keep as is | Keep as is | Brand names are not translated |
Netflix's Hebrew subtitle guide gives a useful default for names going into Hebrew: proper names are transliterated, not translated (its example: Christopher becomes כריסטופר). The same principle works in the other direction.
Caption lines break in the middle of sentences. If each line is translated alone, word order goes wrong, because languages put verbs, negation and adjectives in different places. Translate whole sentences, then split the result back into timed lines. Good tools do this; if you are using a general translator, paste full sentences, not caption lines.
A translation is rarely the same length as the source. Some languages need more characters for the same sentence, and Hebrew, which does not write most vowels, often needs fewer characters than English. Do not assume: check that no line now breaks a reading-speed rule (next section) and re-split long ones.
For anything public, a person who reads the target language natively watches the video with the captions on. Not reads the file: watches the video, because timing errors and a joke that does not land only show in context. If you cannot get a native reviewer for a language, think twice about publishing that language.
Professional subtitling has published rules. Netflix's timed-text style guides are public and specific, so they are a good reference even if your video is a class, not a series. As of 4 November 2026:
| Rule | English | Hebrew | Source |
|---|---|---|---|
| Characters per line | 42 | 42 | Netflix style guides |
| Lines per subtitle | Maximum two | Maximum two | Netflix style guides |
| Reading speed, adult programmes | Up to 20 characters per second | Up to 17 characters per second | Netflix style guides |
| Reading speed, children's programmes | - | Up to 13 characters per second | Netflix Hebrew guide |
| Minimum duration of one subtitle | 20 frames | 20 frames | Netflix timing guidelines |
| Minimum gap between subtitles | 2 frames | 2 frames | Netflix timing guidelines |
Sources: English (USA) Timed Text Style Guide, Hebrew Timed Text Style Guide, Subtitle Timing Guidelines.
What the numbers mean in practice, as arithmetic:
If a fast speaker says a lot in 3 seconds, the translation has to be condensed, not just split. That is a translator's decision, not a tool's, which is another reason for the review step.
These rules are for horizontal video on a TV or a laptop. On vertical clips the screen is narrow, so lines are much shorter (two to five words), and the same reading-speed logic applies.
Hebrew and Arabic captions have problems that left-to-right languages do not:
Translating from English into Hebrew raises the same issues on the output side, and the review should be done by someone who reads Hebrew, on a phone, in the final video.
Translation tools usually charge by minutes of video, sometimes per language. So count minutes, as arithmetic:
Your own time is mostly in step 2 (fixing the source) and step 6 (review). Reading a transcript against the video takes at least as long as the video, often longer. Plan for it, because it is where the quality comes from.
StreamLive Pro transcribes video with a timestamp for every word, then translates whole sentences (not word by word) into 143 languages. You fix the text in a line list beside the video, then either burn captions in with one of more than 80 designs, including a highlight on the spoken word, or export SRT or VTT files to upload to YouTube. Hebrew right to left is handled as a main case: punctuation and numbers land on the correct side. Translation minutes per month depend on the plan and are on the pricing page; the 7-day trial includes 10 translation minutes.
What it does not do: captions or translation that viewers see during the live broadcast are not available yet (in development), and there is no dubbing or voice cloning. More on the captions and translation page, with the steps in Automatic Hebrew subtitles and Caption templates and styles. For clips, see turning a live stream into Shorts.
It translates clear speech well when the source transcript is correct and it works on whole sentences. Names, specialist terms and jokes still need a human check, and anything public should be watched by a native speaker of the target language.
Both work on YouTube. SRT is the most widely accepted and has no styling; VTT is made for web players and supports positioning. If a platform accepts either, SRT is the safer default.
Create an English caption file with the same timing, then in YouTube Studio open the video under Subtitles, choose Add language, pick English and upload the file. YouTube's automatic captions only cover the video's default language.
Netflix's English and Hebrew guides both use 42 characters per line and at most two lines. Reading speed is up to 20 characters per second for English and 17 for Hebrew in adult programmes.
No. It makes translated captions, burned in or as SRT and VTT files, from recordings and uploaded videos. Live captions and translation are in development, and there is no dubbing.
Open vs closed captions: burned into the picture or sent as a file the viewer can switch off. What each is good for, which networks take which, and how to make both.
A weekly Torah class online: live, Shorts and a podcastHow to run a weekly Torah class online: a weekday live, scheduling ahead, Shorts with Hebrew captions, translation for students abroad and a podcast episode.
Burn captions into a video (hardcoded subtitles)Burn captions into a video in StreamLive Pro: correct the text, pick one of more than 80 designs, and render an MP4 with the captions drawn into the picture.
Import SRT captions into CapCut (desktop and web)Import SRT captions into CapCut: make and correct them in StreamLive Pro, download the SRT, and import it on CapCut desktop or web. The phone app cannot.
Want to try it? 7-day trial, no credit card.
Start free