How-to · By Jun Tang · Prices checked November 1, 2026 · ~7 min read

How to Add English + Chinese Bilingual Subtitles to a Short Video with AI

Disclosure: some links on this page are affiliate links (marked "sponsored"). If you sign up through them I may earn a commission at no extra cost to you. It never changes what I recommend — free options are listed whenever they are good enough. More about how this site works.

Bilingual subtitles — English on one line, Chinese (简体中文) on the other — are one of the easiest ways to make a short video travel further. Language learners save them, overseas Chinese audiences share them, and brands selling into both markets get one video instead of two. AI now does most of the heavy lifting, but not all of it. This guide shows the full workflow I use as a short-form editor, which parts AI handles well, and where you still need a human eye.

Why bilingual subtitles work on short-form

Most short-form video is watched with the sound off or low, so captions are already standard. Adding a second language does three extra things:

• Reach. A Chinese-language talking-head video becomes watchable for English speakers, and the reverse. The same upload can perform on TikTok, Reels, YouTube Shorts, Douyin and Xiaohongshu.
• Saves and rewatches. Language learners pause, screenshot and rewatch bilingual clips. Those signals help short-form algorithms.
• Trust for cross-border brands. A product video with correct English and correct Chinese looks professional. A video with machine-translated mistakes looks the opposite — which is why the proofreading step below matters.

Before you start: three decisions

1. Which language is the source?

Always transcribe the language that is actually spoken in the video. Speech-to-text tools are far more accurate on the spoken language than on a translation. For example, quso.ai's subtitle generator states plainly that it "transcribes in the language that's spoken; it doesn't translate" — so you would use it for the first line, then translate separately.

2. Which line goes on top?

Put the language most of your target audience reads on top, in the larger font. For an English-speaking audience learning Chinese, Chinese on top often works better because the characters are the "content" and English is the help. For a Chinese brand selling abroad, English goes on top. There is no single right answer — just be consistent across a series.

3. Burned-in or a subtitle file?

TikTok, Reels and Shorts are built around burned-in (open) captions, so for short-form you will almost always burn them into the video. Keep the .srt files anyway — YouTube accepts uploaded caption tracks, and you will want them if a client asks for edits.

The 6-step workflow

Step 1 — Get a clean transcript in the spoken language

Any of these can produce a timed transcript:

• Free: CapCut's auto-captions, or OpenAI's open-source Whisper model if you are comfortable with the command line.
• quso.ai: its subtitle generator supports 99 languages including Chinese, and paid plans (Lite and up) can download an SRT file. The free plan exports video at 720p with a watermark and does not include SRT download. Cost is 1 credit per minute of video.
• Submagic: markets caption styles "in 123 languages", and is strongest when you want the animated, word-by-word look on top of the transcript.
• ElevenLabs Speech to Text: part of the same credit pool as its voice tools (listed at roughly 330 credits per minute of audio on its pricing page).

Whatever you use, read the transcript once against the audio. Names, brand terms, numbers and slang are where every engine slips.

Step 2 — Split lines for a vertical screen

A 9:16 frame is narrow. Before translating, re-cut long subtitle lines so each one is short: as a rule of thumb, around 30–40 English characters or about 12–16 Chinese characters per line, and no more than one sentence per caption. Doing this before translation keeps both languages aligned on the same timings.

Step 3 — Translate line by line, not as one block

The most common mistake is pasting the whole transcript into a translator and getting back a nicer paragraph that no longer matches your timing. Translate per caption so caption 14 in English is caption 14 in Chinese. If you use an AI chat model, a prompt like this works well:

Translate each numbered line into Simplified Chinese.
Keep the same number of lines and the same numbering.
Keep it short and natural for spoken video subtitles, not formal writing.
Do not translate brand names. Return only the numbered lines.

Some tools have translation built in (Submagic lists "AI Translate captions" among its features; check which plan includes it on the Submagic pricing guide). Built-in translation is convenient, but you still proofread.

Step 4 — Proofread like a human (the step AI can't skip)

Machine translation between English and Chinese is good at grammar and bad at tone. Check for:

• Literal idioms. "That's a game changer" should not become a sentence about games.
• Register. Short-form Chinese is casual; textbook phrasing feels stiff. English captions should sound like speech, too.
• Numbers and units. 万 vs. thousand, prices, dates.
• Names and product terms. Decide once whether a brand stays in English, and keep it that way.
• Length. If a Chinese line is much shorter than the English, that is fine; if the English is far longer than the time on screen allows, shorten it.

Step 5 — Style two lines inside one caption box

Good bilingual captions look like one designed element, not two subtitle tracks fighting each other:

Step 6 — Export and check on a phone

Export at 1080×1920. Then watch the file on an actual phone, not just the editor preview. You are checking that nothing is hidden behind the app's interface and that the Chinese line is readable at arm's length.

Free DIY option: merge two SRT files and burn them in with FFmpeg

If you like free tools and a terminal, you can do the whole thing without any subscription. Once you have en.srt and zh.srt with identical numbering and timings, combine each pair of lines into one caption (English line, line break, Chinese line) to create bi.srt, then burn it in:

ffmpeg -i input.mp4 -vf "subtitles=bi.srt:force_style='FontName=Noto Sans SC,FontSize=14,Outline=2,MarginV=180'" -c:a copy output.mp4

MarginV lifts the captions above the bottom UI area; adjust to taste. This approach is free and precise, but slower to style than a visual editor.

Which tool for which part of the job

If you want to go further than subtitles — an actual Chinese or English voice — read what the ElevenLabs free plan allows before publishing anything commercial.

Common mistakes I fix in client videos

1. Timings drift between languages. Caused by translating as one block. Fix: translate per caption.
2. Captions under the TikTok buttons. Fix: raise the caption box and test on a phone.
3. Traditional vs Simplified mixed up. Mainland China and most learners use Simplified (简体); Taiwan and Hong Kong audiences use Traditional (繁體). Pick one on purpose.
4. Too much text. If viewers need to pause to read, the caption is too long.
5. No proofreading of the source line. A wrong English word becomes a wrong Chinese word, too.

FAQ

Can AI do bilingual subtitles fully automatically?

It can produce a usable first draft in minutes. For anything published under a brand, budget time for a human proofread of both languages.

Is there a free way to do this?

Yes: CapCut or Whisper for transcription, any translator for the second language, and CapCut or FFmpeg to style and export. Paid tools mainly save time and add nicer caption styles.

Should I upload the subtitles as a file instead?

For YouTube long-form, yes — upload SRT tracks so viewers can switch. For TikTok, Reels and Shorts, burned-in captions are the norm.

How long does it take per clip?

For a 30–60 second talking-head clip, about 15–30 minutes including proofreading, once your template is set up.

Related guides