Feature Requests

Currently, when you use a CLM, EVI doesn't start speaking until all the text for its turn has been generated. There is no way for the CLM to start generating some text, have EVI start speaking, and then decide to generate more. This is a feature request to support "streaming" so EVI can start speaking sooner.

The expression API is returning "no healthy upstream" and jobs don't load on the UI.

We are using HUME to teach some phonetics as part of a reading lesson which often requires students to sound out single letters like buh fror B, cuh for C etc. When the learner says these sounds by themselves, Hume does not detect it as a spoken word so it just sits there waiting for a response and then eventually repeats the question. I need it to recognize them saying "buh" as a response and continue to move through the exercise.

Hi, the expression measurement functionality is generally great, but one big hang I have right now is that the transcriptions are pretty rough. Is there any way for me to be able to supply a different transcript as grounding information. (something from whisper_timestamped for example). Thanks!

Hume's audio output quality is phenomenal but I've noticed that the audio input transcription makes quite a lot of mistakes. This has a significant negative impact on UX as it often derails the conversation. When using transcription providers such as deepgram I've managed to configure the options to work very well for my usecase - for example using the model that's specially designed for phone calling and adding custom vocabulary. If we had pass config options to Hume AIs transcription provider then it would enable higher quality transcriptions and significantly improve my users experience.

I would love to see the code implementation required to implement presets. Something OpenAI, Gemini, Anthropic, even Groq does. Currently struggling to piece together how to use a preset with a tool enabled from multiple doc pages & trial and error (which obviously still costs tokens...)

Currently custom_session_id is only passed in message payloads of a websocket. I also need this information during websocket connection.

Presently, there are only two parameters allowed to be passed, config_id and hume_api_key. It would be useful if it accepted additional parameters such as chat_group_id, config_version etc.

This is a feature request (from a Discord user) for the Face model of the Expression Measurement API to return details of the eye color, hair color, skin tone, eye type, and nose shape.