Technology
How Dubline rebuilds the audio
Dubline handles picture and sound separately. The engine works on the dialogue, then joins the new soundtrack to the picture for download.
Picture track
The picture carries through to the output at the source resolution.
Audio track
- Transcribe
- Translate
- Voice
- Mix
The timing of each line guides where the new speech goes.
Inside the dubbing pipeline
Eight stages take the video from upload to download.
-
Read the upload
The engine checks the uploaded video and measures its duration.
-
Extract the audio
The soundtrack is separated from the picture and split into sections for processing.
-
Find the dialogue
Spoken lines are transcribed with start and end times.
-
Translate the lines
The engine translates the dialogue into the selected language using the surrounding lines for context.
-
Synthesize speech
A synthesized voice reads each translated line. Voice selection follows the settings for the job.
-
Fit speech to its slot
Dubline uses the source timestamps to fit and place the new speech.
-
Mix the soundtrack
The new dialogue is mixed with the available background audio.
-
Assemble the MP4
The engine combines the soundtrack and picture, then makes the finished video available in Studio.
Checks along the way
The engine uses the same line timings from transcription through the final mix.
- Khmer script
- For Khmer output, the engine checks the script before sending lines to the voice.
- Timing by line
- Each line has its own time window. The result depends on the speech and the source recording.
- Progress in Studio
- Studio displays job progress so you can see when the video is ready to play.
Hear the result in your own video
Start with a short clip, pick a language, and preview the dub in Studio.