Two languages,
one timeline.
二つの言語、ひとつのタイムライン。
Drop in a video or an audio file. DualCue transcribes the speech, anchors every line to the waveform, then translates inside those boundaries — and hands you back one subtitle file with both languages stacked.
So the hard part was never the model.
難しかったのはモデルそのものじゃなかった。
Drag the track — or flick it and let go
Timing first.
Translation second.
まずタイミング、翻訳はその次。
Speech to timecode
Audio goes to a Whisper-class model that returns every segment with a start and an end. Nothing is guessed from sentence length — the boundaries come from the audio itself.
Lock the boundaries
Each line is pinned to its own slice of the waveform before a single word is translated. The timeline stops being negotiable at this point.
Fill, don't reflow
Lines are translated in numbered batches and checked back against their own numbering. If a line goes missing or gets merged, the batch is rejected and retried — a shifted subtitle is worse than a failed one.
One file.
Both languages.
ひとつのファイルに、両方の言語。
Stacked bilingual output in the three formats that actually get used. ASS carries real styling, so the original sits smaller and greyer under the translation instead of fighting it for attention.
You can also flip the stack order, drop to a single language, or take plain text with timecodes for notes.
The parts that
usually go wrong.
いつも壊れるのは、この辺り。
No cumulative drift
Every cue keeps the timecode it was born with. Hour-long files end as accurately as they start.
Suspect lines are flagged
Silence and noise make speech models invent sentences. Low-confidence segments are marked in the editor instead of quietly shipped.
Line count is verified
Translations come back numbered and are checked line by line. A dropped or merged line fails the batch rather than shifting everything after it.
Click a line, hear it
The transcript is bound to the player. Click any caption to jump there, and the list follows playback — until you scroll, then it gets out of your way.
Japanese, Korean, CJK
Character-based languages are the ones that break naive subtitle tools. They were the first ones tested here, not the last.
Files, not a walled garden
Everything leaves as SRT, VTT, ASS or plain text. Nothing is locked behind a player you have to keep paying for.
Priced by
processing time.
処理した時間に応じた料金。
- 5 minutes of audio per month
- Transcription with timecode
- SRT and VTT export
- One file at a time
- 10 hours of audio per month
- Bilingual translation
- All export formats, including ASS
- Stack order and single-language output
- Email support within 3 business days
- 50 hours of audio per month
- Everything in Pro
- Batch queue
- Priority processing
Prices are in US dollars and exclude any tax that applies in your country.
Plans renew monthly until cancelled. Cancel any time from the customer
portal — cancellation takes effect at the end of the current billing
period and your files stay downloadable until then.
Unused minutes do not roll over. See the
refund policy for the 14-day terms.
Who is behind this. これを作っている人間について。
DualCue is a software product built and operated by DualCue. It exists because subtitle tools kept treating timing as something to fix after translation, which is exactly backwards for Japanese, Korean and Chinese — the languages where a two-second offset stops being cosmetic and starts making a sentence unreadable.
We are a small, focused team. There is no sales department and no chatbot: support email reaches the engineers who build it. Transcription and translation run on third-party model providers, which is stated plainly in the privacy policy along with exactly what is sent to them and what is kept.
Your media is processed to produce your subtitles and nothing else. It is never used to train a model, never sold, and never shared beyond the providers listed in the privacy policy.
- Operated by
- DualCue
- Contact
- support@dualcue.com
- Product
- Subscription software (SaaS) — audio transcription and subtitle translation
- Merchant of record
- Creem
- Service status
- Live — available at dualcue.com/app
A person reads
every email.
メールは全部、人間が読んでいます。
A file that failed, a billing question, a refund, or a language you need that isn't listed yet — write in and you will get a reply within three business days. Include the file name and roughly when you uploaded it and the answer will come faster.