Dubbing replaces the original audio in a file with a new version in a different language, while preserving the voice characteristics and emotional tone of the original speakers. Each speaker is dubbed with a voice selected from the voice library, or with a cloned voice.
Correct speaker assignment is required for dubbing workflows. If imported subtitle files are used to skip automatic audio transcription, ensure all speakers are assigned before proceeding to dubbing.
When two speakers talk at the same time, each speaker needs a separate segment covering the overlapping time range. A segment assigned to only one speaker generates dubbed audio in that speaker's voice alone, even if it contains dialogue from the other speaker.
Use cases
-
A media company wants to distribute an originally English-language documentary in Spain and Japan. The platform's AI identifies each speaker, transcribes and translates the dialogue, and then generates new audio tracks in Spanish and Japanese. The generated voices retain the qualities of the original speakers, creating an authentic viewing experience for the new audiences.
Create a Dubbing
Note
Dubbing can be added only for languages that are already translated. If no language is available, add a translation first.
To create dubbing for a file, follow these steps:
-
In the editor, open the Dub step and click Dub. This step is available only after the Translate step has been completed.
If a speaker has no voice for the dubbing language, the
Voice Assignment Requiredmessage is displayed. Assign the voices in first.The window opens.
-
In , select the required translated languages.
Note
If a regional language variant (e.g., French (Canada), or Portuguese (Brazil)) is selected, the system automatically falls back to the corresponding base language for dubbing.
The regional language selection is still preserved in the project and metadata.
If a selected language already has a dubbing, proceeding overwrites the existing dubbing. The action cannot be undone.
-
Optionally, enable and select existing pronunciations for each language.
-
Optionally, for audio-only recordings, select the option to keep the natural voice speed. Segments grow to fit the voice, so the dubbed audio can be longer than the original.
The uploaded file is stored as the subtitles for the dubbing. The existing translated subtitles are not changed. Dubbing is skipped for the languages with uploaded subtitles.
-
Optionally, upload an .SRT or .VTT file for each selected language in the subtitle upload section to use existing subtitles for the dubbing.
-
Click Start Dubbing.
The window closes. The language drop-down list in the step shows the status of each language.
Note
If a language was dubbed with a different dubbing engine and the current engine does not support it, the window lists the language as not supported.
After the dubbing is finished, use these controls in the step:
-
Dub Changes: Updates the dubbing with the changes made to the text, the speaker settings, or the segment settings. Segments with changes are marked
Edited — re-dub needed, and a counter shows how many segments are waiting. -
Redub This Segment: Regenerates a single segment without dubbing the whole recording again.
-
Re-dub: Dubs all segments of the selected language again. Existing dubbed audio is replaced and cannot be recovered.
-
Sync Dubbing: Appears when the translated subtitles differ from the dubbing. Click it to align the dubbing with the translation.
-
Upload Subtitles: Uploads an .SRT or .VTT file as the subtitles for the dubbing of the selected language, as in step 5.
-
Manage Voices: Selects the voice of each speaker.
Speakers in Dubbing
Each segment is dubbed in the voice assigned to the speaker of that segment. Every speaker needs a voice for the dubbing language. Segments that have no speaker are not dubbed.
If a speaker has no voice, the message One or more speakers don't have a valid voice assigned for this language is displayed with the instruction to open Manage Voices and assign voices before dubbing.
Segments assigned to different speakers that overlap in time are dubbed as overlapping audio. Dialogue that is not captured in a segment does not appear in the dub. To include the dialogue of a second speaker, add a separate segment for that speaker.
Manage Voices
To select a voice for a speaker, follow these steps:
-
In the step, click Manage Voices.
-
Select a speaker in the list. If needed, select the language.
-
Select a voice in one of the tabs:
-
: Voices suggested for the language.
-
: All available voices. Use the search field or the , , and filters to narrow the list.
-
: Cloned, similar, and designed voices of the organization.
-
-
Optionally, click the icon to preview the voice. Without custom text, the voice plays a sample. To hear the voice speak specific text, enter up to 200 characters in the preview field, for example a word with a custom pronunciation. To enter the text in phonetic notation, select Read as IPA (phonetic).
-
Optionally, click the icon to add the voice to Favorites, so that it is listed first in future selections.
-
Click Save.
Note
The available tabs depend on the dubbing engine. The tab is not displayed for organizations that use their own ElevenLabs API key.
applies only to custom preview text. IPA is not supported on ElevenLabs v2, where the text is read as spelled. On Azure TTS the result depends on the locale of the voice and may sound inaccurate.
Dubbing Engine
The dubbing engine determines the available voices and the controls in the editor. Select the engine in , or in the project preferences for a single project.
Available engines:
-
ElevenLabs v2
-
ElevenLabs v3
-
ElevenLabs v4
-
ElevenLabs v4 Turbo
-
Azure TTS
-
Gemini 2.5 Flash TTS
|
Control |
Availability |
|---|---|
|
Audio controls (emotion tags) |
ElevenLabs v3, v4, and v4 Turbo |
|
Segment Volume, Stability, and Similarity |
ElevenLabs engines |
|
Voice cloning and similar voices from the editor |
ElevenLabs dubbing provider |
To use the voices and quota of an organization's own ElevenLabs account, add an ElevenLabs API key as a credential in .
The key requires at least these permissions: Text to Speech (Access) and Voices (Read and Write). Only one ElevenLabs credential per organization can be added.
Audio controls help make dubbing more expressive and accurate by tagging key vocal nuances that add paralinguistic information. Right-click a segment and click to add one or more of the available options at the desired position.
Note
Audio controls are available only with ElevenLabs v3 or v4 engines. Audio controls may not work with all languages or voice models. Testing with the selected voice is recommended to confirm compatibility.
Audio adjustments or redubs after testing do not consume additional minutes.
When a project includes multiple speakers, dubbing settings can be adjusted for each speaker track and for individual segments in the step.
-
Speaker controls
Click the gear icon () next to the speaker name in the waveform to open the settings of the speaker. The settings apply to all segments of that speaker.
-
Segment controls
Select a segment in the timeline to open the panel and adjust dubbing settings as required.
Both panels contain these settings:
-
-
Adjusts how steady or expressive the voice sounds. Lower values make the voice more dynamic and emotional. Higher values make it calmer and more consistent. The labels , , and mark the range.
-
Adjusts how closely the voice follows the selected voice. Higher values sound more like the selected voice. Lower values allow more variation.
Re-dubbing is required after adjusting or . The affected segments are marked Edited — re-dub needed. Click Dub Changes to apply the updated settings.
Adjust Dubbing Speed
By default, dubbing is generated at 1× speed, meaning the system determines the most natural speaking pace based on the amount of text in the segment.
If the text is too long for the time range of the segment, the warning Audio may play faster. Text is too long for the time range. is displayed. The warning appears when the audio would have to play more than 1.15× faster to fit.
To adjust the dubbing speed, use one of these methods:
-
Extend or shorten the speech bubbles on the waveform:
Example
If dubbed audio extends beyond the intended scene at 1× speed, drag the end time of the speech bubble. Dubbing will speed up slightly to fit within the segment.
-
Edit the text and re-dub, or use Shorten Segments over the Limit to get a shorter version of the text.
Example
If audio significantly overflows at 1× speed, the segment may contain too much text. Rewrite the text and click Dub Changes.
Background Audio and Playback
When a dubbing is created for the first time, Phrase Studio separates the speech from the background audio, for example music and sound effects. The dubbed audio is mixed with the background audio, so the original speech is not part of the dub.
To replace the background audio, click Replace Background Audio in the row of the waveform, and upload an MP3, WAV, or M4A file smaller than 300 MB.
Use the slider in the player to balance the original audio against the dubbed audio.
Download the Dubbed Output
To download dubbed output, click Download, select the dubbing language and the content, and click Download again.
The dubbing options are:
-
Dubbed Audio and Dubbed Video.
-
Dubbed SRT and Dubbed VTT.
-
Separated Dubbed Audio: A ZIP file with one MP3 file for each dubbed segment. File names contain the segment number, the speaker name, and the start and end time.
-
Separated Dubbed Audio by Speaker: A ZIP file with one MP3 file for each speaker. Each file contains the dubbed audio of that speaker only, without background audio, on a timeline of the full length of the recording.
Pronunciations control how specific words or phrases are spoken in dubbed audio. They ensure consistent pronunciation of brand names, technical terms, acronyms, and foreign words.
Define custom pronunciation pairs and apply them when creating a dubbing. For existing projects, select pronunciations when adding a new dubbing language in the step of the editor.
Pronunciations are automatically shared with all users of the same organization in read-only mode.
Prerequisites
-
The project's target language supports dubbing.
-
A dubbing language has been added to the project.
To create a pronunciation, follow these steps:
-
On the page, select the tab.
-
Click Create New Pronunciation.
The window is displayed.
-
Enter a and an optional .
-
Select to make the pronunciation available for project selection.
-
Click Save.
The pronunciation is listed in the tab.
Custom pronunciation rules can also be uploaded using the Import button on the tab. Supported files: PLS lexicon (.PLS, .XML), Excel (.XLSX), and CSV (.CSV). Excel files need an "IPA" sheet and CSV files need "Term" and "IPA" columns. CMU notation is not supported.
Note
Save CSV files as "CSV UTF-8", otherwise IPA characters may be corrupted. A file can contain up to 10,000 pairs and be up to 10 MB.
Each pronunciation can include multiple pairs for a single language. To add a pronunciation pair, follow these steps:
-
Select an existing pronunciation and click Create New Pair.
-
In , enter the original word or phrase.
-
Define the desired pronunciation in the field.
Use phonetic spelling, syllable breaks, or approximate sounds.
-
Select the type.
-
Plain Text (spelling)
-
IPA (phonetic)
-
-
Select the relevant language.
-
Optionally, click Preview to listen and adjust as required.
-
Click Save.
The new pair is added to the selected pronunciation.
To add a pronunciation for a word while editing, click Manage Pronunciations in a segment of the step. The pronunciation is saved to the project pronunciations for the selected language. Manage them under .
Existing pronunciations and related pairs can be updated or deleted by selecting Edit next to a pronunciation or a pair.
Re-run dubbing after updating pronunciation pairs; previously rendered audio will not change automatically.
Pronunciation Examples
-
Brand names
-
Apple: ap-pul
-
Microsoft: mai-kroh-soft
-
-
Technical terms
-
GIF: jif
-
SQL: ess-cue-el
-
-
Foreign words
-
Croissant: krwa-san
-
São Paulo: sow-pow-loo
-
Voice cloning generates a synthetic voice based on recordings of a real speaker. The cloned voice can then be used in AI dubbing, allowing translated audio to preserve the tone and vocal characteristics of the original speaker.
The cloning process uses selected speech ranges from uploaded audio or video samples to train the voice model. After samples are processed, a preview can be generated before the voice is saved.
Voice clones are created and managed in the tab of the page. Once created, the voice becomes available for use in dubbing workflows.
Voice sample requirements
-
Total selected range duration must be between 10 and 180 seconds
-
Maximum uploaded file duration: 5 minutes
-
Maximum sample files: 3
-
Maximum file size: 30 MB per file
-
Maximum sample ranges: 50
-
If voice samples contain multiple speakers, ranges must be used to isolate samples from a single speaker.
To create a voice clone, follow these steps:
-
In the tab of the page, click Create New Voice.
-
Select the creation method.
-
-
If is selected, enter a description in the field, and optionally, preview text in the field. Click Generate Previews. Select a voice and continue with the step.
-
Upload Samples to generate the voice clone.
Uploaded media appears in a waveform player where ranges can be marked.
-
Use the timeline to define ranges containing the speaker’s voice, then click Done. Total duration of the sample across all ranges must be at least 10 seconds and at most 180 seconds.
Selected ranges are confirmed. Additional files can be uploaded and processed the same way by clicking Add ranges.
Note
Only the selected ranges are used to generate the voice clone.
-
If the option is used, select the consent checkbox.
-
Click Next.
The step is displayed and the audio sample is generated. This typically takes several seconds.
-
Select to play the preview, and optionally, provide . The new voice speaks every language supported in Dubbing.
-
Click Play Preview or the play icon to preview the new voice.
-
If is selected, choose a voice from generated options. If needed, use Regenerate Previews.
-
Click Next.
The step is displayed.
-
Enter a voice . Optionally, provide voice , and to categorize it.
-
Click Save.
The new voice is added to the list.
Existing voice clones can be previewed, edited, or removed from the tab.
A voice can be restricted to a single project. To restrict the voice, edit the voice and select a Project. Without a project, the voice is available everywhere.
Clone Voices from the Editor
Note
Cloning from the editor is available with the ElevenLabs dubbing provider only.
Create a voice from the speakers of a recording in the step.
-
Click the voice clone icon () in a segment to create a voice from the audio segments of that speaker in the recording. In the confirmation window, select or confirm the right to clone the voice.
-
Click Clone all speakers (N) in the toolbar, where
Nis the number of speakers, to create voices for all speakers. A speaker needs at least 10 seconds of audio in the recording. Speakers with less audio are skipped.
A voice created from the editor is restricted to the project of the recording. To change this, edit the voice in .