LEXI Voice harnesses cutting-edge AI technology to convert live speech into high-quality, real-time translated audio. Built on AI-Media’s flagship LEXI captions, it ensures unparalleled accuracy and fluency in live translation.
Workflow
LEXI Voice fits as a new piece of your current workflow.
LEXI Translate and LEXI Voice
LEXI Voice is an add-on to LEXI Translate, which users enable through an additional LEXI Voice subscription. Ai-Media bills LEXI Voice through a separate line item, and tracks its billing through a separate billing meter. LEXI Voice users may continue to run LEXI Translate without running LEXI Voice.
LEXI Voice subscriptions create additional Voice settings within Translate.
Voice Output with Google Voice
The same engine-agnostic approach that applies to LEXI Live and Translate apply to Voice. Currently, LEXI Voice supports Google Voice:
- Voice Options consist of a mix of Google's Standard and Premium voice options.
- Google Voice labels its options as Language (Dialect) – Gender – Letter
For example, English (Australian) – Female- A. - Selections match their designated output languages, for example, Spanish (Spain) Male A applies to Spanish Translate output.
Supported Languages
Refer to What Languages Does LEXI Voice Support? for a list of supported languages. Language compatibility is subject to expansion with updates to include new voice options and engines.
Voice Models
LEXI Voice users can use voice models to dictate a given term's pronunciation. These models function similarly to LEXI Live's topic models. Voice Models do not require a pre-training stage and Ready confirmation in the same manner that Translate glossaries do. LEXI Voice supports one Voice Model per instance. To create a voice model:
- Navigate to Translate Dashboard -> LEXI Translate Models -> Voice Models.
- On the right-hand side of the Voice Models Manager page, click New Voice Model.
- Enter a Model Name, and select a Language. The voice model's selected language matches the output language of a Voice instance. You can change the model’s language after you create the model.
- Enter optional Tags to help organize your models. Tags do not track model use.
- Click Submit.
Creating Custom Pronunciations
You can use the Pronunciations feature to define phonetic pronunciations for any term you need.
Follow these steps to create a punctuation:
- Enter a Phrase to Pronounce and Phrase to use Instead, as its corresponding pronunciation. Both fields are mandatory.
- Enter phonetic pronunciations in plain-text, as syllable-by-syllable entries in their corresponding alphabet. It is not necessary to use a phonetic alphabet.
- Enter pronunciations in familiar pieces with known pronunciations, and enter a syllable in all capital letters to indicate its emphasis. For example, Henderson as HEN der son or Moreland as MORE land.
- You can identify entries as case sensitive. For example, this prevents the system from mispronouncing the last name Sessions and the common word sessions.
- Multi-word entries will not overlap the pronunciation of individual words. For example, if you define a custom pronunciation of the name Tom Brady, the pronunciations of the individual words Tom and Brady retain their common pronunciation.
- Click Add Phrase, and add another phrase if necessary.
- Click Upload Pronunciations or Download Pronunciations to import or export your pronunciation dictionaries as .csv files. This is similar to LEXI Topic Models or Translate DNT and Glossary models.
Note: As with other forms of topic models, take note if the system converted any characters during the upload or download processes. - Click Save Changes to complete your additions.
Voice to Encoder
LEXI Translate generates output to LEXI Voice, on the same access code or service as Translate’s text output. In other words, wherever text goes, Voice goes. For example, a Translate instance configured for Access Code 3 and targeting S3 sends its text and Voice output to the designated access code and service.
Support for Lexi Voice is currently available with iCap Alta.