Voice Management

In this paradigm, you can manually associate a specific speaker with each dialogue key via its metadata.

Although SAD mode is primarily designed for one actor per key, it is possible to associate multiple voice profiles with the same key via metadata, for example, for generic messages requiring different intonations or voices.

Example of a dialogue managed in

SAD

mode with two distinct keys:

text
hi, how are you?

text
yes, I'm great!

In this paradigm, multiple speakers can intervene within the same dialogue key.

Their lines are identified by specific tags that segment the text.

The order of intervention is sequential.

To extract character IDs and their associated texts, you must define a regular expression (regex).

In some complex projects, it may be judicious to group several dialogues into a single key and manage their separation in post-production, directly within the game engine.

Example for a

MAD

dialogue:

text
{lia} hi how are you?
{boo} yes I'm great
{lia} Oh!
{boo} what!? why that face?
{lia} ...
{sam} ...<br/>hey you two!

This method significantly reduces the number of keys to manage.

For the example above, where each line would have its own key, and with a translation into 10 languages, this would represent more than 60 keys.

Managing each intervention individually would quickly become unmanageable for an RPG containing hundreds of such interactions.

List of all audio instances created for the selected key.

These tabs group characters by language.

List of entities (characters) to which a voice narration can be assigned.

In "MAD" mode, their order corresponds to their appearance in the source text.

The captured text serves as the basis for the generated voice narration.

It is recommended to modify punctuation or add emotional cues to influence the voice, without altering the original source text.

An alert will be displayed if the source text is modified.

This button allows you to enhance the text intended for voice narration based on the character's personality and the context defined in the key's metadata.

This feature is only available for ElevenLabs V3 models.

This button generates a voice and adds it to the history.

This button undoes changes made to the text and restores the original captured version.

Allows you to validate a voice.

Validated voices are kept independently of the history and will be exported according to the criteria defined when saving the project.

It is possible to validate multiple voices for the same text and character.

File names will then be automatically incremented.

This can be useful, for example, to nuance voices in a game engine using random seeds.

This information, used during generation, serves as a signature to track and understand the origin of each voice.

For example, if you add a new character or move them within the source text, you will be able to reassign existing voices to the entity, even if its location has changed.

Allows you to play the voice and visualize its pitch to compare it with the desired overall mood.