Translate a clip: the same clip in another language

1How to translate a clip

1. Open the clip

   In Studio open a finished clip full screen — in the “My generations” feed or in a project.

2. Press “Translate the clip”

   In the Actions bar at the top of the viewer. The “Translate the clip” window opens.

3. Choose the language

   The “Into which language” block has 9 options: English, Russian, Spanish, Indonesian, Arabic, German, French, Turkish, Portuguese. You do not choose the language of the original speech — it is detected automatically.

4. Choose the voice

   In the “Voice” block click the row you want: your clones first (marked “my voice”), then the catalogue voices. More in the voice section below.

5. Decide about subtitles

   The Burn the translated subtitles into the video box puts the translation on the picture in large white text with an outline. Without the box the sound is new and the picture stays as it was.

6. Check the price and press “Translate”

   The “Price” line shows the number of tokens, and under it is “Clip length N s. If there is no speech or something fails, the payment is refunded”. The button shows the same number: “Translate · N tokens”. If you tick “Burn the translated subtitles into the video”, the price rises a little (on short clips this can add 1 token to the number), and a note appears under the price: “Burned-in subtitles are a separate part of the price: if they can’t be burned in, it isn’t charged”.

While the translation runs, the button says “Translating…”, and the window does not close with Esc or a click outside. When it is done, the window says “Done. The clip with new voice-over was added to your feed” and offers the buttons Open the clip and .srt. The original clip stays as it was, and the new one appears in the feed as a separate entry labelled “Translation · language” (shown in Russian as “Перевод · язык”).

2What happens inside

• Speech recognition. Everything that is spoken in the clip is found; the language is detected automatically.
• Translation. All lines are translated in one request, the number of lines is kept and the answer is checked. The translation is in a lively spoken style, and each line is kept close to the original length.
• Voice-over. Each phrase is voiced with the chosen voice. One voice for the whole clip: if several people speak in it, they will all speak with the same voice.
• Assembly. The phrases are placed at their time slots. If the voice-over is longer than the original phrase, it is sped up, but by no more than 1.4 times. The picture of the clip is not re-encoded and loses no quality; only the sound changes.
• Subtitles (optional). The translated text is burned into the picture.

3Voice: the catalogue or your own clone

• Catalogue voices. These are paid ElevenLabs voices — up to 12 in the list: male (for example “Брайан” — Brian, “Ксавьер” — Xavier, “Джеймс” — James) and female (for example “Белла” — Bella, “Аделина” — Adelina, “Ева” — Eva); the names are shown in Cyrillic in every interface language. Next to each it says male or female. The translation price does not depend on which one you choose.
• My voice. If you have a ready voice clone, it stands first in the list marked “my voice” and is selected by default. You can clone only your own voice — how to make a clone is on the voice cloning page.
• You cannot pick someone else’s clone. Other people’s voices are not in the list, and the server rejects a clone that is not yours or does not exist.

A hint in the window reminds you: “Only your own voice can be cloned — other people’s voices aren’t in the list”.

4Subtitles and the .srt file

If you tick “Burn the translated subtitles into the video”, the translation is placed on the picture in large letters with an outline — the same look as “Large with outline” in the subtitles window.

The .srt button in the result window downloads a translation.srt file with the translated lines: you can upload it to YouTube or open it in a video editor. It is available even if you did not tick the subtitles box.

5How much it costs

The price has several parts: speech recognition (by clip length), voice-over of the translation (by the number of characters, at most 20 characters per second of the clip and 2500 characters overall), translating the text (a fixed $0.02) and, if “Burn the translated subtitles into the video” is ticked, a separate $0.01 share for the subtitles (about 0.4 of a token). Studio calculates the price from the length of your file, takes the box into account and shows the sum before you start.

6Limits and restrictions

A translation takes a place among your tasks in progress: while it runs, it counts as one of the three you can hold at once when starting new clips.

If you exceed the run limit you will see “Too many requests in a row. Try a little later”; if the clip is longer than 2 minutes — “For translation the clip must be at most 2 minutes”; if the clip is not yours — “Only your own clip from Studio can be translated”.

7What is not here yet

• No lip sync. The lip movement in the frame does not change: a clip with a person talking will look like a dub. Lip sync is not available yet.
• Background music and noises are not kept. The original sound is replaced entirely, only the new voice-over remains.
• Clips longer than 2 minutes are not translated.
• One voice for the whole clip: there are no different voices for different speakers.
• The language of the original speech cannot be chosen by hand, it is detected automatically.
• You cannot translate someone else’s clip by link, only your own from Studio.

Related