How to Translate a Video: Subtitles, Voice-Over and Transcripts

Step 1: decide what kind of translation you need

Every video translation starts with the same step — turning speech into text — but ends in different places. Pick the output that fits how people will watch:

Translated subtitles

Best when viewers can read along, for social feeds watched on mute, and for keeping the original speaker's voice.

Translated voice-over (dubbing)

Best for long videos, tutorials where eyes are on the action, children, and viewers who prefer not to read.

Translated transcript

Best for understanding, research, quoting and documentation — no one has to watch the video at all.

If you're unsure, compare them in subtitles vs dubbing. mydubly produces all three in one run when you switch Full translation output on.

Step 2: prepare the file

• Use the original export rather than a re-downloaded or screen-recorded copy — cleaner audio means a more accurate transcript, which means a more accurate translation.
• Trim silence, intros and outros you don't need translated; translation is priced per minute.
• Check the length: mydubly handles up to 2 hours per file. Split longer recordings.
• MP4, MOV, WebM, MKV and M4V all work. See MP4 and MOV notes if your file is unusual.

Step 3: translate it

1. Open the video translator and drop your file in. The video stays on your device — your browser extracts only the audio.
2. Choose the target language. You don't need to pick the spoken language; it's detected automatically.
3. For a voice track, switch Full translation output on and pick one of 8 voices. For subtitles and transcripts only, switch it off — it costs 1 credit per minute instead of 50.
4. Wait while the audio is transcribed, translated and voiced in 30-second chunks. Keep the tab open.
5. Download what you need: the translated MP4, the translated audio, SRT/VTT subtitles and both transcripts.

Step 4: review before you publish

Machine translation is fast and usually accurate for clear, everyday speech, but it can stumble on names, jargon, idioms and humour. Before publishing:

• Read the translated transcript, or ask a native speaker to skim it — 10 minutes of review catches most issues.
• Check names, brand terms and numbers.
• Watch the first minute and a random middle section of the translated MP4.

Step 5: publish

Upload subtitles as separate tracks (YouTube, Vimeo, LinkedIn and most course platforms accept SRT or VTT — see how to add subtitles to a video). Publish a dubbed MP4 as its own video, or as an extra audio track where the platform supports it.

Match the video to a mode before you upload

The steps above describe three outputs, but in the tool you choose one of two modes, and each mode returns a fixed set of files. Settling the mode first saves you from paying for a voice track nobody will hear, or from running the job twice because you needed a file the other mode produces.

Dub mode

Transcribed, translated, read aloud by one stock AI voice and merged with your original picture. Downloads: translated video, translated audio (M4A), the original-language transcript, and SRT and VTT files that match the spoken translation.

Transcript mode

Transcribed and, if you pick a language other than the spoken one, translated as text with no audio. Downloads: timestamped transcript, SRT, VTT and a plain transcript (translated when you chose another language, untranslated when you chose the spoken language).

Price

Dub mode is 50 credits per minute with a 2-minute minimum per file; transcript mode is 1 credit per minute with a 5-credit minimum per file.

A simple rule: if nobody will ever listen to the translation, use transcript mode. That covers subtitles, research, internal review, and checking what a foreign-language clip says before deciding whether it deserves a dub at all.

Which video types suit which mode

• Screen-recorded software demo with one narrator: dub mode. Viewers watch the cursor rather than a subtitle line, and a single stock voice suits a single narrator.
• Two-person interview filmed on camera: transcript mode with translated subtitles. A dubbed version would read both people in the same voice, and viewers lose track of who is talking.
• Short social clip that autoplays without sound: translated subtitles, since a voice track goes unheard while the feed is muted.
• All-hands or lecture recording over an hour: either can work. If you dub it, trim the Q&A or breaks you don't need, because every minute is billed.
• Foreign-language footage you only need to understand: transcript mode with your own language as the target, then read the result screen, where the original and translated text sit side by side.

Worked example: one webinar, two languages

A team has a 20-minute product webinar in English. They want a dubbed Spanish version and German subtitles.

1. Cut the 3-minute waiting-room intro in an editor first, leaving 17 minutes. That alone saves 150 credits (15¢) on the dub.
2. Run the 17-minute file in dub mode with Spanish as the target. For viewers in Mexico or Colombia, choose Spanish (Latin America) rather than Spanish (Spain) in the picker; the Spanish video translator page covers that choice. Cost: 850 credits (85¢).
3. Run the same file in transcript mode with German as the target and download the SRT. Cost: 17 credits (1.7¢).
4. Keep the original-language transcript from the Spanish run. It is the reference a reviewer checks both translations against.

When the output isn't what you expected

• You got an MKV instead of an MP4: the original video codec can't be stored in an MP4, so the in-browser merge used MKV. VLC and most editors open it; re-export from an editor if your platform insists on MP4.
• The dub-mode subtitles differ from a transcript-mode translation of the same video: in dub mode, subtitles follow what the voice actually says, including lines shortened to fit timing. Pair the dub-mode SRT with the dubbed video and the transcript-mode SRT with the original audio.
• Some passages come out in the wrong language: the spoken language is detected automatically, and a video that switches languages partway can be handled inconsistently. Split the file where the language changes and run each part on its own.
• Names and product terms are wrong in the translation: recognition errors flow straight into the translation, so search the original transcript first. If the name was misheard there, the translation never had a chance. See translating names and technical terms.
• Slide titles and captions in the picture stay untranslated: only speech is translated. Plan replacements as described in translating text that appears in the picture.

Reviewing a translation without watching it twice

The result screen pairs each original line with its translation, so a bilingual reviewer can work down one view instead of juggling two documents. Ask them to check three things in order: numbers and units, names, and sentences whose meaning flipped, because a dropped negation is a classic machine translation slip. Then play the dub preview at 1.25× or 1.5× to hear the whole track quickly, slowing to 0.75× wherever a line sounds rushed. The error patterns worth hunting for are listed in common machine translation errors, and translating a video into multiple languages explains how to keep terms consistent when you go beyond one language.