Feature Requests

In my AI film platform, users design voices that can be reused across different films. A common use case is designing a voice for an English-language film, then wanting to reuse that same voice in a Spanish-language production. Currently, this isn't possible because voices are tightly bound to the original design language. Request: Enable designed voices to be multilingual for use in other supported languages (e.g., Spanish), preserving the same vocal identity and tone. This would greatly enhance voice reuse and consistency across multilingual projects.

We love Hume's emotionally expressive TTS, especially with instant_mode. But for next-gen AI applications — like animated avatars and voice assistants — we'd love to see support for true real-time WebSocket streaming. This would allow developers to send text and receive playable audio in tiny chunks (≤1s) with ultra-low latency, enabling lifelike back-and-forth conversations without waiting for full MP3 rendering. Current HTTP-based streaming works well, but a socket-based approach would take Hume to the next level for live interaction and animation sync. Please consider this!

Would love a way to integrate into our existing mobile app to enable TTS locally. We have a hard requirement to be offline capable as users are using our app outdoors and won't necessarily have access to a reliable internet connection.

Octave voices are currently only supported for Hume's Text-To-Speech API . Soon, we plan to add support for Octave voices to the Empathic Voice Interface , so you will be able to use them in conversations with users. Follow this post for updates!

Why is it not letting me do a child voice

It would be amazing if you could add a RAG support directly on the platform for adding specific knowledge for the EVI. Thanks!

Prebuilt integrations to use EVI with no-code tools like Zapier, Make.com , Pipedream, and n8n. Community question: Please let us know what tools are most important for you!

When using an EVI with a custom language model there is no way the limit the chat history. Long conversation sessions result message payloads that overwhelm AWS API Gateway max message size for websocket. It should be possible to specify a maximum chat history length. We manage history ourselves in our custom LLM, so we would set this max length value to 0. The expectation is that it would only send the latest user input with each new message.

Currently, only being able to update the settings of a live session/connection over the frontend client means making the prompts used visible to the user. If we have sensitive business logic related prompts and actions we use for steering the hume conversation we might want to apply those without the user client knowing. This could be solved with some form of hook or proxy, I am not sure. But it would greatly benefit bigger production projects that rely on their business logic being not easy to instantly copy.

This is a feature request (based on what people have written into Discord) to allow clients to connect to an EVI conversation via WebRTC, (as opposed to websockets). WebRTC removes some of the guesswork with respect to audio formats, and has some nice "adaptive" features that let the bitrate of the audio adjust to connection quality, I believe it also has built-in support for echo cancellation.