Character Training

Chatting in Studio​

Before sending them out into the world, we suggest chatting with your character to see if they’re what you expected. Are they the right combination of personality and knowledge? Or should you tweak something? Refining a character can take time to get it just right. But we think it’s definitely worth the effort!

Chat with characters directly in their profile by clicking “Chat” in the top right. This will open a pop-up window where you can type messages back and forth with your character. Alternatively, press the microphone icon for a live voice chat.

Explore the menu buttons to expose additional options for moving to a scene, triggering goals, and restarting or exporting the current conversation.

Regenerate​

While chatting with the character, if you’re unsatisfied with their response, click the “Regenerate” button below their speech bubble. This will cause the character to generate a new response, giving you the opportunity to right any deviations in the dialogue. Repeat this function until you’re presented with the response you envisioned. Click the “Versions” button to review previous regenerated responses and select the best one before continuing the conversation.

Rate​

Give the character’s responses thumbs up or thumbs down to help direct future answers. This rating provides information about which response you prefer, which helps inform the character about optimal responses going forward.

Chat Logs​

In the workspace’s left sidebar, select Interactions to access previous conversations for characters within that workspace. This is a way to see the interactions that your users have had with the characters. And in turn, inform opportunities for improvements to the character’s responses and align with the intended behaviors.

For safety purposes, we anonymize all users and only allow analytics access after a character has been shared a certain number of times. Interaction data is also only currently available to people who have Professional or Custom accounts.

LLMs​

Is Metamorph run off of Open AI’s LLMs (GPT-4, ChatGPT)?​

We have 20+ models that drive our characters. This includes non-LLM models like emotions, text-to-speech, speech-to-text, and more. For LLMs, we dynamically switch between LLM APIs and our own proprietary models depending on which is best for the conversational context and latency. Our goal is to provide you with the best and fastest answers. We understand the strengths of the individual LLM APIs and our own models, and dynamically query the one that will provide the best character experience.