New persona-2 model is now available for image generation. It is the next-gen persona model with improved identity preservation and higher fidelity output. Appearance params (gender, body_type, skin_color, auto_detect_hair_color) are ignored by this model. Available in /image/v1/generate_gallery_image and /chatbot/v3/response endpoints.
The instruct model has been removed from the /chatbot/v3/response endpoint, as the roleplay model now provides the same level of instruction-following capabilities.
New supreme model is now available for image generation. It delivers hyper-realistic results with superior text rendering and minimal artifacts. Available in /image/v1/generate_gallery_image and /chatbot/v3/response endpoints.
The base model has been removed. All requests now default to large model with improved quality. Existing API calls with "model_name": "base" will continue to work (automatically converted to large).
New recipe available! See details here https://docs.exh.ai/recipes/animate-image#/ 🔥
instruct - a model optimized for precise instruction following and controlled task execution within roleplay scenarios.
Increased maximum context length in /chatbot/v3/response from 20 to 40.
Added support for the new persona model in the model_name field in /image/v1/generate_gallery_image endpoint enabling image generation with significantly improved consistency in appearance and style relative to the provided reference image.
Removed the example_messages field from user_profile and updated name to be optional.
Added example_messages field to bot_profile body parameter in the https://api.exh.ai/chatbot/v3/response endpoint. This optional list parameter enables specifying example messages to guide the model in generating responses that capture the character's personality and style.
