Two-way live translation
Translates live speech in both directions so each participant can speak their own language and hear the other in theirs.
Eleven Realtime Translation
Eleven Realtime Translation converts live speech into natural translated audio for two-way conversations across more than 90 languages. It is intended for telecom, contact center, event, and other multilingual communication workflows, with selectable voices and streamed captions or transcripts.
Live speech translation for multilingual conversations
Eleven Realtime Translation is a live speech translation product for conversations in which participants speak different languages. It converts speech into translated audio as people talk, allowing each person to speak in their own language and hear the other in theirs.
The product is designed for latency-sensitive, two-way interactions rather than prerecorded dubbing. It supports more than 90 output languages, automatic language detection, selectable voices, and streamed translated text with source-audio transcripts. ElevenLabs presents it for carrier networks, contact centers, multilingual events, and other global conversation workflows.
Teams can tune how quickly translation begins through the eagerness setting and configure audio formats through an API. The service is described as enterprise-ready, with SOC 2, PCI DSS Level 1, ISO 27001, HIPAA, GDPR, and Zero Retention Mode listed on the product page; implementation details, product-specific limits, and pricing for Realtime Translation are not specified in the supplied material.
Core features
90+ language output
Supports output in more than 90 languages while maintaining a consistent selected voice across the conversation.
Automatic language detection
Automatically detects the language each caller speaks, reducing the need for callers to select a language manually.
Selectable voice layer
Lets teams choose from a library of more than 17,500 voices or use an Instant or Professional Voice Clone. The selected voice is used for translated speech rather than reproducing the original speaker exactly.
Captions and transcripts
Streams translated text together with interim and final transcripts of the source audio, supporting captions and logging workflows.
Configurable responsiveness and audio
Provides an eagerness setting with low, medium, and high options to trade translation completeness against response speed. Audio formats can be configured for telecom use through the API.
Practical use cases
Live events
Conference organizers can stream translated speech to attendees’ devices during conferences, press briefings, and multilingual meetings.
Telecom services
Telecom operators can add live translation to network conversations so subscribers can communicate across languages without switching to a separate translation workflow.
Contact centers and BPOs
Contact centers and business process outsourcers can let customers and agents converse in their native languages while serving customers across markets.
Multilingual communication products
Product teams can build multilingual calling or conversation experiences around streamed translated audio, captions, and transcripts through configurable API-connected audio workflows.
Pros and Cons
Pros
- Supports bidirectional live translation for natural cross-language conversations.
- Offers output in more than 90 languages with automatic language detection.
- Provides voice selection, including a library of more than 17,500 voices and Instant or Professional Voice Clones.
- Streams translated text and interim and final source transcripts alongside audio.
- Lists enterprise security and privacy options including SOC 2, PCI DSS Level 1, ISO 27001, HIPAA, GDPR, and Zero Retention Mode.
Cons
- The product does not preserve the original speaker’s voice exactly; translated speech uses a selected library voice or voice clone.
- The supplied product material does not specify Realtime Translation pricing, setup requirements, or detailed language-by-language limitations.
- Two-way behavior and endpoint configuration require implementation at the network or product level.
FAQ
Can it translate both sides of a conversation?
Yes. The product supports bidirectional conversations: each participant speaks their own language and hears the other participant in theirs. Bidirectional behavior is configured at the network or product level, such as by connecting each caller to a dedicated endpoint instance.
Does it provide captions or transcripts?
Each session streams translated text along with interim and final transcripts of the source audio. These outputs can support live captions or session logging in addition to translated audio.
Does the translated audio preserve the original speaker’s voice?
Realtime Translation uses a selected output voice rather than carrying the original speaker’s voice across languages exactly. Users can choose from ElevenLabs’ library of more than 17,500 voices or use an Instant or Professional Voice Clone as the voice ID.
How can teams balance translation speed and quality?
The eagerness setting controls the balance between speed and translation quality. High starts translation sooner for lower latency, low waits for more of the sentence for potentially higher-quality output with more delay, and medium is the default.
Which speech model produces the translated audio?
Eleven v4 Turbo is the default text-to-speech model for translated speech. The source also states that customers can set the TTS model to eleven_flash_v2_5 or eleven_v3.
Quick Facts
- Category
- AI speech translation
- Primary workflow
- Live, bidirectional conversation
- Language coverage
- 90+ output languages
- Audio output
- Natural translated speech with a selectable voice
- Text output
- Translated text plus interim and final source transcripts
- Provider
- ElevenLabs