# AI Dubbing (/docs/ai-services/ai-dubbing)



AI Dubbing turns the speaker's voice into spoken audio in another language, live. Your audience can listen in their own language instead of reading captions.

  AI Dubbing is in **beta** and is switched on per organisation. [Contact
  us](/contact) to enable it. In the dashboard it appears as
  **Interpretation**, under AI Agents in your project settings.

## How it works [#how-it-works]

AI Dubbing uses a **speech-to-speech** model: the original audio goes in, translated speech comes out, with no captions in between.

We currently use one model: **Google Gemini Live Translate**. It works on a continuous stream, translating while the speaker is talking rather than waiting for them to finish a sentence.

* **Language changes**: the model follows a live stream that switches language, without restarting.
* **Voice matching**: the dubbed voice follows the original speaker, including their gender.
* **In sync**: every dubbed track stays aligned with the original video.

## Supported languages [#supported-languages]

AI Dubbing is available in 75 languages, listed below and on the [Languages](/languages) page. The source must be a single-language input: inputs that combine several languages can't be dubbed yet.

| Code | Language |
| --- | --- |
| AF | Afrikaans |
| SQ | Albanian |
| AM | Amharic |
| AR | Arabic |
| HY | Armenian |
| AZ | Azerbaijani |
| EU | Basque |
| BE | Belarusian |
| BN | Bengali |
| BG | Bulgarian |
| CA | Catalan |
| ZH | Chinese (Simplified) |
| HR | Croatian |
| CS | Czech |
| DA | Danish |
| NL | Dutch |
| EN | English |
| ET | Estonian |
| FI | Finnish |
| FR | French |
| GL | Galician |
| KA | Georgian |
| DE | German |
| EL | Greek |
| GU | Gujarati |
| HA | Hausa |
| HE | Hebrew |
| HI | Hindi |
| HU | Hungarian |
| IS | Icelandic |
| ID | Indonesian |
| IT | Italian |
| JA | Japanese |
| JV | Javanese |
| KN | Kannada |
| KK | Kazakh |
| KM | Khmer |
| RW | Kinyarwanda |
| KO | Korean |
| LO | Lao |
| LV | Latvian |
| LT | Lithuanian |
| MK | Macedonian |
| MS | Malay |
| ML | Malayalam |
| MR | Marathi |
| MN | Mongolian |
| MY | Myanmar (Burmese) |
| NE | Nepali |
| NO | Norwegian |
| FA | Persian |
| PL | Polish |
| PT | Portuguese |
| PA | Punjabi |
| RO | Romanian |
| RU | Russian |
| SR | Serbian |
| SD | Sindhi |
| SI | Sinhala (Sinhalese) |
| SK | Slovak |
| SL | Slovenian |
| ES | Spanish |
| SU | Sundanese |
| SW | Swahili |
| SV | Swedish |
| TL | Tagalog (Filipino) |
| TA | Tamil |
| TE | Telugu |
| TH | Thai |
| TR | Turkish |
| UK | Ukrainian |
| UR | Urdu |
| UZ | Uzbek |
| VI | Vietnamese |
| ZU | Zulu |

## Adding AI Dubbing [#adding-ai-dubbing]

1. Open your project's **Settings** and find **Interpretation** under **AI Agents**.
2. Click **+** to add an interpretation.
3. Choose the **Model** (Gemini Live Translate).
4. Choose the **Input source**: the language input whose speech you want to dub.
5. Choose the language to dub into, then click **Add Interpretation**.

The interpretation shows as **Idle** until audio is flowing, and you can add one for each language your audience needs.

## Where the dubbed audio goes [#where-the-dubbed-audio-goes]

| Destination                                         | How the dubbing is delivered                                                                                        |
| --------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------- |
| [HLS streaming](/docs/destinations/hls-streaming)   | Each dubbed language is published as an alternative audio track that viewers pick in the player.                    |
| [RTMP streaming](/docs/destinations/rtmp-streaming) | The original audio is replaced with one dubbed track. Turn off **Use original audio** and pick the **Audio track**. |
| SRT                                                 | Dubbed audio from existing interpretations can be muxed into the SRT output.                                        |
| Captions page                                       | Coming soon: listen to the dubbed audio on the same page as the captions.                                           |

## Things to know [#things-to-know]

The model's voice replication is good but not perfect. As Google notes, a voice can occasionally shift after a long pause, start with the wrong gender, or settle on one voice when several people talk in quick succession. Language detection can also struggle with heavy accents or very similar languages, although the translation itself should still be accurate.

For a high-stakes session, test with a recording of your speakers first, and keep captions available alongside the dubbed audio.

Dubbed audio is processed by the same pipeline as the rest of your content: see [Data handling & retention](/docs/security/data-handling).