Translate video captions and keep the timeline | AI Video Factory

The usual jam when taking a video overseas is not “finding a translator.” It is exporting captions into a second project: the clock is gone, bilingual lines slip, and one edit means realigning the whole file. The stabler path in the studio: review the source captions, then translate on the same timeline. Timestamps stay put. Stack the translation or keep it alone.

Clean source cues first

A script with no timestamps can only be translated as one block — length is a guess. The right order:

1. Get source-language captions with in/out points (speech, burned-in text, or an imported SRT).
2. Fix typos, names and breaths, then open translation.
3. Translation changes the words only. Do not start a second project.

The ten-minute caption path is in this guide. Translating a messy timeline only copies the mess into another language.

Translation on the same clock

1. Open the studio with an existing project or drop in a local MP4 or MOV. The file stays in the browser.
2. Confirm the source cues are reviewed. If there is no timeline yet, recognize speech or import SRT first.
3. Tick Translate captions and pick one target language. One language per run — Japanese and English means two passes.
4. To show source and translation together, turn on Bilingual captions. On vertical video, keep the translation larger; landscape can stack two lines.
5. Export a hard-sub MP4, or save SRT / VTT for this language.

The free plan includes one translation run per month. Languages and limits are on Features and Pricing.

Review only three things

• Names: people, brands and model numbers get paraphrased first. Skim against the source before you publish.
• Line length: translations often run longer than Chinese. Split a vertical line past a short clause; do not cover faces.
• Breaths: translation does not move timestamps. If one cue overflows, edit that line or its in/out — do not rerun the whole file.

What usually breaks

Pasting the whole script into a generic translator and sliding it back by hand: lengths never match, bilingual lines drift. Clean the clock first, then translate. Clips with many proper nouns or mixed accents still need a human pass — the model is for speed, you own the last mile.

If you still need a foreign-language voice, translate first, then synthesize on the same timeline — see AI dubbing. The full post order is in creator post in 2026.