Dubbing

Dubbing on Morphic translates a video or audio recording into another language and speaks it back in each original speaker's own voice. Morphic transcribes the speech, translates it, and re-voices it with the speakers' delivery and timing, so the dub lines up with the original.

What comes back depends on the model. Morphic Dub returns the video itself, with each speaker's close-ups lip-synced to the new speech. ElevenLabs Dubbing returns a new audio file, even when you start from a video.

When to use itCopy Anchor Link

• Release a finished episode, ad or course in another market without generating it again
• Keep a recognizable voice, such as a host or a recurring character, across languages
• Make one version per market from the same master

For a new read in a different voice, use Speech. For another voice in the same language, use Speech to speech. For translated captions instead of new audio, use Transcription.

Choose a modelCopy Anchor Link

Pick Morphic Dub for a video. Pick ElevenLabs Dubbing for an audio file, a video longer than 10 minutes, or on the Free plan. Both take your own SRT transcript and translation instead of transcribing and translating themselves: Morphic Dub with assigned voices, ElevenLabs Dubbing in the original voices with the spoken language set; see The Dubbing settings panel. Morphic Dub is slow: a run takes up to about 13 minutes per minute of video, and a short clip takes longer than that ratio suggests, since its fixed stages cost the same whatever the length.

Dub from the prompt barCopy Anchor Link

1. Open a file on Morphic

2. In the prompt bar at the bottom, open the mode picker and choose 'Audio', then 'Dubbing'

   

3. Pick the model in the prompt bar: 'Morphic Dub' or 'ElevenLabs Dubbing'. Paid plans start on Morphic Dub

   

4. Add one file under 'References': upload it, or pick it from your Canvas or Assets. Morphic Dub takes a video; ElevenLabs Dubbing takes an audio or video file

   

5. Click the settings icon in the prompt bar to open the Dubbing settings, then choose a language under 'Dub into'. The list is searchable. ElevenLabs Dubbing includes regional versions, such as 'American English' and 'Arabic (Egypt)'; Morphic Dub dubs a regional version in its main language

   

   

6. (Optional) Leave 'Spoken language' on 'Detect automatically', or set it when the recording is noisy or mixes languages; with a transcript of your own on ElevenLabs Dubbing it must be set. Under 'Keep as spoken', type the names and terms that must not be translated, comma-separated

7. Click the blue button in the prompt bar. It shows the credit cost, worked out from the length of your recording

The dub lands on your Canvas, the same length as the source: a video from Morphic Dub, an audio file from ElevenLabs Dubbing. While it's made, a card in its place shows 'Creating...' and the time left.

The Dubbing settings panelCopy Anchor Link

The settings icon in the prompt bar opens the Dubbing settings. With Morphic Dub selected it holds everything a dub can take:

With 'Assign a voice to each speaker', a speaker row appears for each person you cast. Name the speaker by a face image or a voice clip attached to the prompt, picked under 'Who speaks', or describe them under 'Describe the speaker', such as 'The man in the red jacket'; then choose their voice under 'Pick a voice' from Morphic's voice catalog. 'Add a speaker' adds a row. Anyone you do not name gets a fitting catalog voice. The 'Transcript' and 'Translation' pickers list the SRT documents attached to the prompt; attach the file first to pick it there.

On Morphic Dub, a transcript or translation of your own needs assigned voices: with the original voices, the dub is made in one pass that takes no text, so the pickers stay off. See Options for what each option does and when to use it.

With ElevenLabs Dubbing selected, the panel holds 'Dub into', 'Spoken language', 'Keep as spoken', 'Transcript' and 'Translation'; there is no 'Voices' setting, since every speaker keeps their own voice. 'Transcript' takes your own SRT of the speech in the spoken language, and 'Spoken language' must then be set, since the engine cannot detect it from a transcript. When most cues start with a speaker label, such as Mira: …, the labels decide who speaks when, one voice per speaker; without them everyone is dubbed as one speaker. 'Translation' turns on once a transcript is picked and takes your own SRT of the lines in the target language: every transcript cue needs a translated cue over it, since the translation is timed by the transcript's cues, and 'Keep as spoken' still applies whenever ElevenLabs translates. Cues must lie inside the recording; a cue longer than 25 seconds is split.

Dub with CopilotCopy Anchor Link

1. Open Copilot

2. Attach the recording with the paperclip

3. Ask for the dub and name the language, for example "Dub this into Spanish"

   

On a paid plan, Copilot dubs a video on Morphic Dub and tells you the expected wait. An audio file, a video longer than 10 minutes and every dub on the Free plan go to ElevenLabs Dubbing; from a video, Copilot then puts the dubbed audio back over the picture, and the lips do not follow the new speech. To choose yourself, name the model in the request.

What you can add to the same request:

• Several languages: "Dub this into Spanish, French and German" runs one dub per language, side by side
• The source language: name it when the recording is noisy or mixes languages. Otherwise it is detected
• Names and terms: list character names, places or product terms that must not be translated, for example "Keep the names Mira and Halden as they are". Without them, names can come out translated literally
• A voice per speaker: @-mention a frame or a clip of the speaker and the voice, for example "Dub this into Spanish. Give @Image 1 the voice @Alice and pick fitting voices for the rest". Speakers you do not cast get a fitting catalog voice; Copilot picks from Morphic's voice catalog, not from your organization's own voices
• Your own transcript or translation: attach the SRT with the paperclip and say which it is, for example "Dub this into Spanish with assigned voices, using @interview.en.srt as the transcript" or "speak the lines in @interview.es.srt". On Morphic Dub both need assigned voices; Copilot tells you so rather than dropping the file. On ElevenLabs Dubbing, name the model and the spoken language, for example "Dub this into Spanish on ElevenLabs Dubbing; it is spoken in English, use @interview.en.srt as the transcript"; a translation goes with its transcript

Dub over MCPCopy Anchor Link

With the MCP server connected, ask your assistant in the same words, or have it call dub_media with the clip's media_asset_id, the target_language as a code such as hi, the names as keyterms and, to choose the model, model_name: "Morphic Dub" or model_name: "ElevenLabs Dubbing". Your own SRT goes as source_transcript_asset_id or translation_asset_id; on ElevenLabs Dubbing they need source_language beside them, and the translation goes with the transcript. Each language is one call. The call returns a generation the assistant waits on.

Dub over the Direct APICopy Anchor Link

POST /v1/direct/tools/dub_media takes the same fields, with the model as model_id: "mdl_h2z5a7af7y50", and returns a generation to poll or to receive by webhook. Morphic Dub charges an estimate at start and settles it to the run's spend. The full guide, with curl, Node.js and Python examples, is Dubbing over the Direct API.

LimitsCopy Anchor Link

What it costsCopy Anchor Link

Dubbing is charged on the length of the source recording, so a 2-minute clip costs the same in every language. Each language is charged separately. Morphic Dub costs more per minute than ElevenLabs Dubbing. The cost shows on the blue button before you run it, and current rates are on Plans & Credits.

On Morphic Dub the cost on the button is an estimate: it is charged when the dub starts and settled to what the run actually spent when it finishes, never more than three times the estimate. A dub that fails is refunded.

The whole file is charged, so to test voices and names, trim a 20-second sample and dub that first.

TroubleshootingCopy Anchor Link

Copilot asks which languageCopy Anchor Link

Name the target language in the request. Copilot never guesses it from the language of the chat.

The dub is a sound file, not a videoCopy Anchor Link

The dub ran on ElevenLabs Dubbing, which always returns audio. For a dubbed video with synced lips, pick Morphic Dub on a paid plan, with a video up to 10 minutes long. To keep the audio dub, ask Copilot to put it back on the video, or line it up with the clip in Compose.

A name was translatedCopy Anchor Link

Run it again and list the names and terms to keep: under 'Keep as spoken' on the prompt bar, or by telling Copilot the names in the request.

A speaker's face matches nobodyCopy Anchor Link

The frame you attached for a speaker shows nobody the dub found on screen, so their voice was not assigned and they were voiced like the others. Attach a clear, front-facing frame from the video itself, one face per image, or name the speaker in words instead ('the woman in the red dress').

Set the spoken languageCopy Anchor Link

A transcript of your own was given to ElevenLabs Dubbing with the spoken language left on detect, which the engine cannot do from a transcript. Set it and run again: 'Spoken language' in the Dubbing settings, the language in the Copilot request, or source_language in the call.

The SRT is in the wrong languageCopy Anchor Link

The transcript is in the spoken language and the translation in the target language; a file in the other language is dubbed as it is, so the result speaks the wrong language or the lines land off their cues. Check which picker, or which field, you gave the file to, and give the two files different names so they do not get swapped.

The dub is taking longCopy Anchor Link

A Morphic Dub run takes up to about 13 minutes per minute of video, and the fixed stages make a short clip slower than that ratio: a 50-second clip took 17 minutes. The card on the Canvas shows the time left; Copilot tells you the expected wait when it starts the dub.

That's how you use Dubbing on Morphic. Reach for it when a finished recording needs to reach another language in the same voices.

---

If you have any trouble or have any questions, feel free to reach out. We're happy to help. Write to us at support@morphic.com.