This is a guide to the AI text and voice generation features of the web editor. It enables users to create rich, interactive dialogues with customizable text and voice outputs, supporting multiple languages and voice profiles.

Drop the script named “phrases” to the element and the features will be available in the inspector

Character Description: (optional) Allows users to input detailed descriptions of characters within the experience. The AI model uses the provided character description to generate text that aligns with the character's traits, background, and personality.

Prompt field: It serves as the main input for generating text. Users can input questions or prompts that the AI will respond to, considering the character's description. It works in conjunction with the character description to generate contextually appropriate text responses.

Example: “Elon Musk” is sufficient enough character description for the AI model 😅. Generally, for characters in the experience, verbose descriptions help the model to come with better answers

Same input does not guarantee same output - it uses OpenAI’s API in the background.

Direct text generation is also possible in many other languages. Choose the appropriate languages from the dropdown

Text field: This field will generally be populated by the AI generated text after clicking on “GENERATE TEXT” button. You can edit this field once it has been populated by the AI generated text. Also, if you already have the dialogue ready to be converted to speech, the prompt field may be empty. The value in the text field will be used to generate audio/speech.

Speech generation: Clicking on “GENERATE SPEECH” will convert whatever text is available in the text field to speech in the voice of the selected voice-profile. The speech will be generated in whatever language the text is in - it is simply text to speech for the selected voice profile. Note that at this stage, the audio asset is not yet part of the asset library. Keep experimenting with different texts and voice profiles until you finalize a dialogue.

Voice Profile dropdown: The user may choose one of the available voice profile present in the dropdown menu - the profiles are provided by Eleven Labs. Please contact R&D if you need to add a custom voice profile

SAMPLE USER FLOW:

Screen Recording 2024-07-08 at 5.07.23 PM.mov

Multiple Phrases Since a character will have many dialogues - multiple phrases are supported. Clicking on the “+” button next to phrase selector will create a new phrase with a “default” name. You can update the name in the Phrase Name field as per convenience. Note that two phrases cannot have same names - in case when a user enters similar name, a random suffix is automatically added to the phrase name in order to differentiate it from the other similar name

Example of multiple phrases and switching between them