How can I ensure my voice isnt trained in the principal model from Elevenlabs? so they produce one

[primal pelican]

AI-generated Answer for relay_stevan_12568

ElevenLabs does not use the data you upload to train their main models. Your voice clone will not be shared unless you explicitly enable the option to share your voice. If you want to ensure your voice is not used to improve the services, you can request to opt out by completing this online form. Once processed, your voice will no longer be used for service improvements, but this does not affect any prior uses.

If you wish to delete your voice, you can do so by navigating to your Voice Library, selecting the voice, and using the delete option. If you encounter any issues, you can reach out for assistance.

[north juniper]

Elevenlabs produce a model file per voice ?

[primal pelican]

AI-generated Answer for relay_stevan_12568

ElevenLabs does not produce a separate model file for each voice. Instead, voices are generated using their existing models, such as the ElevenLabs Voice Models. If you want to ensure that your voice is not used to improve these models, you can request to opt out. You can do this by completing an online form to opt out of your Content and User Voice Models being used to improve the Services, including the ElevenLabs Models. Once your request is processed, your voice will no longer be used for this purpose, but it won't affect any prior uses. For more details, you can refer to the ElevenLabs Terms of Service.

[north juniper]

but how elevenlabs will remove my voice if elevenlabs trained their principal model using the data?

[primal pelican]

AI-generated Answer for relay_stevan_12568

ElevenLabs allows you to opt out of having your data used to improve their AI models. To do this, you can modify your data use settings by clicking your profile icon on the ElevenLabs platform, navigating to “Terms and privacy” > “Data use,” and disabling the toggle for 'Improve the models for everyone.' This ensures that new data you submit won't be used to train their models.

If you have previously submitted an opt-out request, your choice should be reflected in your Data use settings. For more detailed information, you can refer to the ElevenLabs Terms of Service and their data use policy.

Once you opt out, your voice data will no longer be used to improve the models, but this does not affect any uses of your data prior to the opt-out request being processed.

[north juniper]

Elevenlabs create a model pth file per voice or use as inference?

[primal pelican]

AI-generated Answer for relay_stevan_12568

ElevenLabs does not create a separate model file for each voice. Instead, they use a set of foundational models to generate audio outputs. If you wish to opt out of having your voice data used to improve these models, you can request this by completing an online form. Once processed, your voice data will no longer be used for model improvement, but this does not affect any prior uses of your data. For more details, you can refer to the ElevenLabs Terms of Service.

[north juniper]

explain better about set of foundational models

[primal pelican]

AI-generated Answer for relay_stevan_12568

ElevenLabs offers two main families of foundational models: standard models and Turbo models. Each family includes both English-only and multilingual models, designed for specific use cases with varying strengths in speed, accuracy, and language diversity.

1. Standard Models: These are optimized for quality and accuracy, making them ideal for content creation like voiceovers and audiobooks. They include:

   • Multilingual v2: Offers high stability and accuracy in 29 languages, best for lifelike speech.
   • English v1: The original model, now surpassed by newer models.
   • Multilingual v1: An experimental model, now outclassed by newer versions.

2. Turbo Models: These are designed for low-latency applications, such as real-time conversational AI. They include:

   • Turbo v2.5: Supports 32 languages with low latency, optimized for real-time interactions.
   • Turbo v2: An English-only model focused on speed for conversational applications.

Each model is trained differently to cater to specific needs, whether it's high-quality content creation or fast, real-time interactions. For more detailed information, you can refer to the Models page.

[north juniper]

So when I request to train my voice using PVC elevenlabs use their principal model or create a separate model file exclusive to my voice?

[primal pelican]

AI-generated Answer for relay_stevan_12568

When you request to train your voice using Professional Voice Cloning (PVC) at ElevenLabs, a separate, custom model is created exclusively for your voice. This involves training a dedicated model on a large set of your voice data to produce a hyper-realistic clone that is indistinguishable from the original voice. This process ensures that your voice is not part of a general or principal model but is instead used to create a unique model file specific to your voice.

[north juniper]

all voices from elevenlabs is separated model file or a principal model using as inference?

[primal pelican]

AI-generated Answer for relay_stevan_12568

ElevenLabs uses a principal model for inference rather than creating separate model files for each voice. When you use the Professional Voice Cloning feature, your voice data is used to create a custom AI model that works within the framework of the existing principal model. This approach allows for efficient processing and high-quality voice synthesis without the need for individual model files for each voice.

[north juniper]

Explain better regarding how this architecture works

[primal pelican]

AI-generated Answer for relay_stevan_12568

ElevenLabs uses foundational models that are trained on a vast amount of audio data. These models are designed to handle a wide variety of tasks, from natural conversational speech to more dramatic audiobook readings. When you create a voice using the Professional Voice Clone feature, it doesn't create a separate model file exclusive to your voice. Instead, your voice data is used to fine-tune the existing foundational models to generate outputs that sound like your voice.

The architecture relies on these foundational models to perform inference, meaning they use the trained data to generate new audio outputs based on the input provided. This approach allows for efficient use of resources and ensures that the models can handle a wide range of voices and styles without needing to create a separate model for each voice.

If you wish to remove your voice data from being used to improve these models, you can request to opt out. This means your voice data will no longer be used for model improvement, but it won't affect any outputs or models that were created before your request was processed.

[north juniper]

and how elevenlabs will remove my voice from a trained model?

[primal pelican]

AI-generated Answer for relay_stevan_12568

ElevenLabs uses foundational models to generate synthetic audio, which means they do not create a separate model file for each individual voice. Instead, they use a principal model that can generate various voices based on the input data provided.

When you create a voice model, such as a Professional Voice Clone, it is not a separate model file but rather a configuration or set of parameters that the foundational model uses to generate audio that sounds like your voice.

If you request to remove your voice from the system, ElevenLabs will stop using your data to improve their models. However, any improvements or changes made to the foundational model using your data before the removal request will remain. The removal process ensures that your data is no longer used for future model training or improvements.

[north juniper]

which kind of parameters?

[primal pelican]

AI-generated Answer for relay_stevan_12568

I'm sorry, but I don't have enough information to answer your question about the specific architecture and parameters used by ElevenLabs for their voice models. For detailed inquiries about how voices are managed or removed from models, I recommend reaching out directly to ElevenLabs through their ticket form for assistance.

[north juniper]

so elevenlabs will use my voice to train the model multilingual or they will train a specific model for my voice?

[primal pelican]

AI-generated Answer for relay_stevan_12568

When you use ElevenLabs' Professional Voice Cloning, a dedicated model is created specifically for your voice. This process involves training a separate model using a large set of voice data to ensure a hyper-realistic representation of your voice. This means your voice is not used to train the general or principal models but is instead used to create a unique model file exclusive to your voice.

[north juniper]

elevenlabs create a model for each voice or they use a sample as reference?