Articles in this section

Creating LEXI Live Instances

Updated

LEXI Live is a live captioning service that you control through EEG Cloud. This document describes creating a new LEXI Live instance. LEXI Live instances apply automatic captions to iCap access codes and have settings that control the captions. These settings include the on-screen placement of captions and management of questionable language.

 

Additional Documentation

Refer to the following articles, which this document references: 

Before you Begin

You must have an EEG Cloud account to configure LEXI Live. If you do not have an account or an active LEXI Live subscription, contact the EEG sales team at sales@eegent.com or call 516-293-7472, option 2.

Be sure to configure iCap on your encoder before you use LEXI Live. See iCap™ Networking and Firewall Requirements for more information. 

Creating a New Instance

To create a new LEXI Live Instance:

  1. Sign in to EEG Cloud with your Cloud Services username and password, and then click LEXI Live. The LEXI Instances page opens and lists all of your current instances. 
  2. On the LEXI Instances page, click Add New.
  3. Complete the fields in the Create A New LEXI Instance window. Fields marked with an asterisk (*) are required. The table below describes all settings.
  4. Optionally click Advanced Options to display additional settings to further configure your new instance. These settings display on a new page, which carries over your selections from the basic settings page. You can find these additional settings listed below the Advanced Options row in the table.

  5. Click Submit. Your new instance displays on the LEXI Translate Instances Page.

    Setting Name Description
    * indicates required fields
    Name *

    Enter a descriptive name for your instance.

    Example: French Broadcasting
    Tags

    Add labels to help organize your instances. 

    Enter tags one at a time, and then click the tag's bubble. For example, French. 


    Note: You must click tag's bubble to add the tag.
    Access Code *

    Enter the iCap access code corresponding to the encoder that receives the translated captions. For more information, see How Do I Create an Access Code?

    Note: You must share your access code with the iCap company EEGASR for it to be available to a LEXI instance.
    Engine * Provides the language model for this access code's captions. This setting changes the supported languages.
    Speaker Change Style

    Select how LEXI indicates when there has been a change in speaker.  For more information, see  Selecting Speaker Change Styles for detailed descriptions of these options. 

    Options are:

    • Off: Default, displays no indicator.
    • Chevron (>>): Displays a double chevron as an indicator.
    • Dash (-): Displays a dash as an indicator.
    • Color Change (W/Y): Uses two alternating colors to indicate speaker changes: white for the first speaker and yellow for the second.
    • Color Change (W/Y/C/G): Uses four colors to indicate changes between consecutive speakers: white for the first speaker, yellow for the second, cyan for the third and green for the fourth.
    • Numeric (Speaker #): Displays numeric speaker indicators in sequential order (Speaker 1, Speaker 2, etc.).
    Language

    Select the broadcast's language. The supported languages change depending on the Engine you select. 

    Note: Some models include untranslated words from other languages. For instance, the French word fromage appears in English models. 

    For a list of all supported languages, visit What Languages Does LEXI Live Support?
    Topic Model Select a topic model from your list of models. 

    Note: Some models include untranslated words from other languages. For instance, the French word fromage appears in English models. 

    For information about topic models, see LEXI Topic Model Setup.

    Advanced Options

    Displays additional configuration options in an expanded window, which carries over your settings for the basic options (above). The following settings in this table are the advanced options. The starred options are mandatory only if using the advanced options.

    Output Mode *

    Select an output mode that aligns with your caption encoder's CC Output Format setting and the international region where your content will be viewed. The option you select here changes which of the CC Service and Teletext Page options display.

    Options are:

    • 608/708
    • Newfor/Teletext
    CC Service *

    Available only with Output Mode 608/708 selected.

    Select the CTA-608 caption channel to send captions to. Generally, this is primary language (CC1/S1). Use S1 for CC1/S1 and S2 for CC3/S2. 

    • Primary language (CC1/S1)
    • Secondary language (CC3/S2) options.
    Teletext Page *

    Available only with Output Mode Newfor/Teletext selected.

    Enter the magazine and page number. Match this setting to your encoder's Teletext Page setting.

    By default, encoders configured for OP-47/Teletext output use a Teletext Page setting of 801 for the primary caption channel.
    Number of Rows The number of lines of captions displayed. Options are 2, 3, and 4 rows.
    Base Row *

    Select the base/bottom row for captioning. This field represents the CTA-608 base row or the Teletext base row depending on the Output Mode you select:

    • With 608/708 selected, rows 2 (top of the screen) through 15 (bottom of the screen) are available.
    • With Newfor/Teletext selected, rows 3 (top of the screen) through 22 (bottom of the screen) are available.
    Column Width The number of characters per text line.
    All Caps Toggles whether your captions are in mixed case (unchecked, default) or all capital letters (checked).
    Erase the screen when LEXI enters standby or is stopped Disables the display when LEXI is not running.
    Profanity Filter Selects a profanity filter from your account to use with this instance. Profanity filters remove objectionable language from your broadcasts' captions. These filters replace unwelcome language with other text, (i.e. a more suitable word, or nonsense such as *@K!@#$X), or remove it entirely. LEXI Live enables these filters by default. For more information, see Creating Profanity Filters.
    Max Delay(s)

    The maximum number of seconds between receiving audio input and producing CC output. Higher max_delay values may yield greater recognition accuracy. Default is 3.5 seconds.

    This value represents the maximum time that LEXI takes to process audio and return results. Overall end-to-end latency, from audio input to CC output, is determined by a user's workflow (encoder, network traffic, decoder, etc.). 
    Disfluency Filter When enabled, filters out filler words such as “um”, “uh”, “ah”, etc.
    Display Style * Sets the caption-advance style. Available options are Roll-Up and Pop-On.
    Column Indent * The number of horizontal character positions between the margin and the start of text.
    Inactivity Timeout *

    LEXI will terminate if it has not produced output for the selected amount of time. Select None to configure LEXI Live to never time out.

    Default: 10 minutes.
    Enable Audio Events

    Use this option to select the types of audio events that LEXI identifies with text captions. For example, music displays as the word Music. Available events are:

    • Music
    • Applause
    • Laughter 

    Note: LEXI may misidentify applause or laughter as the other. If you encounter this, disable one and watch your captions for improved accuracy.
    Enable Multi-Track

    Select this option if you plan to uplink multiple tracks of speaker-isolated audio from your EEG encoder. In these cases, you can assign a label here to each separate speaker track. LEXI will listen for the number of tracks specified and use the provided labels in track order. 

    Note: AI-Media bills this feature with an extra 50% of your basic rate per additional track. Do not enable this feature without first enabling multi-track audio uplink on your encoder.
    Use LEXI Vision to automatically position captions Select this option to configure LEXI Vision uses AI to set captioning screen position automatically based on your video content, avoiding text and faces. LEXI Vision overrides your fixed row position settings.

Selecting Speaker Change Styles

LEXI Live offers these text styles to indicate when there is a speaker change. These styles comply with local caption standards. 

  • Off (no speaker changes): LEXI uses no speaker change indicator.
  • Chevron (>>): Displays a double-chevron when the speaker changes.
  • Dash (-): Displays a dash when the speaker changes.

  • Color Change (W/Y): Captions alternate between white and yellow text to indicate each change of speaker, Regardless of the number of speakers, this alternating pattern indicates a new speaker. For example:

    • First speaker: white
    • Second speaker: yellow
    • Third speaker: white
    • Fourth speaker: yellow
    • etc.

  • Color Change (W/Y/C/G): LEXI assigns colors based on the number of consecutive, unique speaker changes, while tracking a rolling sequence of up to four consecutive unique speakers. These colors do not indicate a specific speaker; they indicate a relative position in the sequence. The same speaker might be in white at a given moment, in yellow later, and then green if they return as the fourth speaker in the sequence. 

    • Two speakers, indicators change alternate white and yellow. This is effectively the same behavior as the Color Change (W/Y) option.
    • If discussion rotates between three speakers, indicators change from white, to yellow, and then to cyan.
    • If discussion rotates between four speakers, indicators rotates between white, yellow, cyan, and then green, in that order.

    For example, imagine a newscast with a panel discussion. Indicators are:

    1. Anchor (first speaker): white.
    2. Field reporter (second speaker): yellow.
    3. Return to the anchor (first speaker): white.
    4. Meteorologist (third speaker): yellow. The meteorologist's speech is in yellow because yellow always follows white (the anchor).
    5. The anchor (first speaker) introduces a three-speaker panel: 
      • Anchor (first speaker): white. 
      • Panelist A (fourth speaker): yellow. This panelist's speech is in yellow because yellow always immediately follows white.
      • Panelist B (fifth speaker): cyan. New speakers who follow speakers in yellow are always in cyan.
      • Panelist C (sixth speaker): green. New speakers who follow speakers in cyan are always in green.
    6. Return to the Anchor (first speaker): white.
    7. Sportscaster (seventh speaker): yellow.
  • Numeric ID: The system displays numeric speaker indicators in order of appearance (Speaker 1, Speaker 2, etc.). Numeric speaker labels display as Speaker #: [text], for example, Speaker 12: I yield the remainder of my time, with each speaker on a new line. 
    Notes:
    • Speaker-to-number assignments remain consistent over the course of a LEXI job. For example, Speaker 1 remains Speaker 1 every time they speak. 
    • Sequence and assignments reset with each new activation of a LEXI instance. 
    • LEXI uses the English term Speaker, regardless of the output language you select. LEXI maintains the numeric formatting LEXI Text to LEXI Translate output.

 

 

 

 

Related to