Short answerOpen Operations → Content → Languages and choose Upload finished episodes (or Tools → Drama Dubbing), create a project for the drama, upload an episode (MP4 or MOV, up to 5 minutes and 500 MB), check who is speaking, then pick languages. Each language comes back as a dubbed MP4 plus an SRT file. One unit is 80 credits and covers up to 80 seconds of lip-synced dialogue; you see the count before you pay, and failed jobs are refunded.
Drama Dubbing is for episodes that were not made in Dawnwood: a finished drama you shot, bought or produced elsewhere, or a title you license and hold dubbing rights to. For episodes made here, order language versions from Operations → Content → Languages or from the Review page instead (see the language versions guide).
Uploaded dramas sit in Operations → Content → Languages next to the series you made here, so you can see every series and language in one place.
Drop in an MP4 or MOV, up to 5 minutes and 500 MB per file. Longer files are refused; split them into episodes first. The file needs an audio track. You can drop several files at once: episode numbers are read from the file names (ep03, E03, S01E05), and files without one take the next free number.
Processing runs by itself and shows its stage on the episode row:
Nothing is charged for this part.
When processing finishes, the episode asks you to check the speakers before it can be translated. Each speaker has a short sample you can play.
Why we ask: detection works from the sound alone, and two similar voices (two older women, say, or two men of the same age) can be merged into one speaker. If that happens, both characters would be dubbed in one voice. A minute spent here prevents that.
From the second episode on, the speakers found are matched to the ones already in the project. Matches the tool is unsure of, and speakers it thinks are new, are flagged for you to confirm.
Dubbing is counted in units of 80 credits per episode per language.
Example: a 40-episode season into Brazilian Portuguese, one unit each, is 3,200 credits.
Each language appears as its own row on the episode. When it is ready you can preview it and download:
Each speaker is dubbed in a voice cloned from their own lines in the footage. If a speaker has too little clean speech to clone (under about two seconds), a stand-in voice of the same gender is used and the speaker is marked stand-in.
If lip-sync fails on some shots, those shots keep the original mouth movement, the result is marked as partial, and the credits for that part are refunded. The rest of the episode is delivered as normal.
Producing the same language again for an episode bills again.
If a translation fails, its row shows the failure and the unused credits go back automatically to the pool they came from. If it fails after lip-sync has started, the lip-sync cost already spent is kept at cost and the rest is refunded. If an upload fails during processing, retry it from the episode row; processing is free, so a retry costs nothing.
Voices are cloned only for your job and deleted when it finishes. Your uploads and results stay in your account, private to you. You own the dubbed output, and your uploads are not used to train models.
To remove an episode, delete it from its row. To remove everything, delete the project: its episodes, speakers, uploads and results are removed with it, including the stored files. While a translation is running, neither its episode nor the project can be deleted; wait for it to finish or fail.
Still stuck? Email support@dawnwood.ai with your account email, the series, the episode and the segment or group — see what to include.