How to Translate a Long Video with AI Voice-Over: Lectures, Courses and Podcasts up to 2.5 Hours

Your best video is probably already finished: a recorded workshop, a course module, a long interview, last season of your podcast. Putting it in front of Spanish, German or Japanese viewers used to mean a transcript, a translator, a voice actor and an afternoon in an audio editor — for every episode.

The video translator on ImageGPT now has a long-form mode for exactly this. Upload your own video from 1 minute to 2.5 hours, review the translated script, pick the voices, and download an MP4 with the new language next to the original. You see the price of every step before you start it.

What the result looks like

We took the opening 2 minutes 14 seconds of “Tour of the Moon”, a narrated flyover built from Lunar Reconnaissance Orbiter data by NASA's Scientific Visualization Studio. NASA video is public domain, so it's fair game. We uploaded it, picked Spanish and the voice-over mode, and left every other setting on its default.

The AI found one speaker, split the narration into 18 lines and translated them as connected sentences rather than word by word — “Orientale basin” became “la cuenca Oriental”, while the names LRO, Diviner and LOLA were left alone. From upload to finished video took just over a minute.

The finished MP4 keeps your original picture untouched — no re-encoding. Next to it sit two audio tracks, the translation (default) and the original, plus subtitles in both languages. Players like VLC let viewers switch between them, and you can also download the subtitles as SRT and VTT files for YouTube or your course platform.

What you can translate

Only content you own or have the rights to: your own videos and channel episodes, courses and webinars, interviews, podcasts, documentaries you hold a license for. You confirm this with a checkbox before anything runs. Movies and TV shows you don't own are off limits.

• Length from 1 minute to 2.5 hours, files up to 6 GB.
• MP4, MOV, MKV or WEBM with H.264, H.265, AV1 or VP9 video. If your file uses something else, convert it to MP4 first in the free video converter.
• 18 languages in any direction: English, Spanish, French, German, Italian, Portuguese, Polish, Turkish, Dutch, Ukrainian, Russian, Japanese, Korean, Chinese, Arabic, Hindi, Indonesian and Vietnamese.
• Clean speech transcribes best. Loud music over the voice or people talking over each other means more lines to fix in the script.

Step by step

1. Open the video translator and switch to “Film, lecture, episode — up to 2.5 h”.
2. Pick a project: a new one, or an existing one if this is the next lesson or episode. Choose the language to translate to; leave the original language on “Detect”.
3. Choose the output: Voice-over, Full dub or Subtitles only. For a voice-over, choose the voices: one narrator, a male and a female voice matched to each speaker, or a voice per character.
4. Upload the file or paste a link to it. Large files go up in parts, so if your connection drops, pick the same file again and the upload resumes where it stopped.
5. Confirm you have the rights to the video and press “Transcribe and translate”. The price is on the button.
6. Review the cast, the glossary and the script. Every line has play buttons for the original and the translation, and you can edit any line.
7. For a long video, press “Try the first … min” first: it voices only the opening, up to two minutes past the first line, so you can hear the voices before paying for the rest.
8. Press “Voice the whole video”, then “Download video” and, if you need them, “Subtitles”.

Everything runs on our servers, so you can close the tab and come back to a progress bar with the time left. Stop a render halfway and the credits for the unvoiced part come back. The finished video is also saved to My works.

Voice-over, full dub or subtitles?

A full dub separates the voices from the music and sound effects with Demucs, then lays the new speech over the original background. Each character can get a stock voice, a voice you saved on one of your characters, or “Voice from the video” — a copy of that speaker's own voice speaking the new language. Copy only your own voice or the voice of someone who agreed to it; you confirm that with a separate checkbox. Voice copies aren't available for Ukrainian, Indonesian and Vietnamese yet, so those get a stock voice.

Translating a whole course or podcast season

Put every episode of a series into one project. The project remembers the cast — up to 12 named speakers with their voices, with everyone else grouped as “Other men” and “Other women” — and a glossary of names and terms. In the next episode the AI recognizes returning speakers by their voice, so your co-host sounds the same in lesson 1 and lesson 10.

Fill in the glossary after the first episode. If a cooking course translated “fold in” two different ways, add the term with the translation you want and press “Translate again”: only the translation is redone, and lines you edited by hand stay as they are.

How much it costs

You pay per step. Transcription and translation are charged right after the upload, voicing after you've reviewed the script. The price depends on the length and on how much talking there is, and every button shows the exact amount. Rough numbers for a talk-heavy video:

Our two-minute NASA clip came to about 20 credits, transcription included. Videos with long pauses or music sections cost less, since only spoken lines are voiced. Fix a single line later and you pay for that line only; rebuilding the video re-mixes just the parts that changed.

Short clips under 5 minutes still go through the regular mode of the same tool, which adds lip sync and burned-in subtitles. For animated captions on a Reel or a Short, use Auto Captions.