LLM

This panel gives you control over your digital assistant's intelligence by connecting a Large Language Model (LLM), which serves as the brain behind your AI avatar's conversational abilities. This configuration directly impacts how naturally, accurately, and effectively your avatar communicates with users.

Out of the box, Genesis Studio offers a selection of industry-leading LLMs as part of its integrated services:

• Gemini: Google's advanced language model, available directly in the Studio.
• ChatGPT: OpenAI's versatile conversational AI; you can use the built-in model or supply your OpenAI API key to access custom or fine-tuned models from your own account.
• Azure OpenAI: Microsoft-hosted access to OpenAI models with enterprise controls, available only by connecting your own Azure resource via API.

After making your choice, click Save below to apply the updated configuration.

Tip:  

The list of supported LLMs will continue to expand over time. If you'd like to use a different LLM for your project, you can contact the RAVATAR team at support@ravatar.com to discuss possible integration. Please note that custom setups involving third-party services are priced individually based on scope.

Linking Custom OpenAI GPT Model

While the standard options deliver strong performance, you can also create a more tailored setup by connecting your own OpenAI GPT model via API. This enables your digital human to speak in your brand’s voice, reflect domain-specific knowledge, and stay aligned with your project’s purpose and role in user-facing scenarios.

To do this, choose ChatGPT from the list of supported LLMs and tick the 'Use own OpenAI API key' option below.

In the field that appears, enter your API key and click Apply.

Note:   Using your GPT model via API typically incurs additional charges from OpenAI based on token usage.

Once connected, the system performs a handshake with your specified LLM and retrieves a comprehensive list of available models, including base models, fine-tuned versions, and any additional modules.

Select the Model you require from the list that appears (you can use the built-in search field for quick navigation – e.g., type mini to locate the gpt-4o-mini model), and click Save to apply the new project configuration.

You can now talk to your enhanced AI avatar in the Live Preview panel to evaluate the changes. In case you encounter any issues with its interactivity, please refer to the instructions on how to choose the right custom GPT model.

Tip:   You can switch back to one of the integrated LLMs at any time by deselecting the 'Use own OpenAI API key' option and saving your project.

How to Choose the Right Custom GPT Model

If your AI avatar remains silent after being linked to your custom OpenAI LLM, it's likely you've selected the wrong option from the retrieved list of supported models.

Here are a few tips to help you make the correct choice:

• For custom fine-tuned models, OpenAI assigns names following the ft-{identifier} format, where 'identifier' is an automatically generated model ID. These models usually appear at the bottom of the model selection list and do not retain base model labels (such as 'gpt-4') in their names.

• If you haven't fine-tuned a custom model yet, or if you'd prefer to use a different GPT version than the one provided in Genesis AI Avatar Studio by default, you can choose any base GPT model by referencing its official OpenAI name (e.g., gpt-4o-mini, gpt-4-turbo, text-davinci-003 for GPT-3.5, and so on).

After selecting the appropriate model, make sure to click Save at the bottom of the wizard before addressing your AI avatar to verify the changes.

If you're still experiencing issues with your AI avatar's responsiveness despite following the recommendations above, feel free to contact the RAVATAR team at support@ravatar.com or use the Support option in the dashboard (accessible via the envelope icon in the top-right corner). We'll get back to you as soon as possible to help ensure your AI avatar is set up and functioning properly.