Feature Requests

I'm getting an API error with no error message. https://github.com/HumeAI/hume-typescript-sdk/issues/712#issuecomment-3884958129 This is the url encoded string: access_token=DHo4HAIo3dGcthgrKu08QAtITgJK&config_id=7e315a6f-4230-4a67-b128-bbe567bcb96b&verbose_transcription=true&session_settings%5Bsystem_prompt%5D=You%20are%20an%20AI%20voice%20agent%20for%20a%20law%20firm.%20Your%20mission%3A%20collect%20essential%20information%20from%20callers%20efficiently%20while%20creating%20a%20professional%2C%20human%20experience.%0A%0A%3Cidentity%3E%0A-%20You%20are%20an%20intake%20specialist.%20Your%20job%20is%20to%20gather%20caller%20information%20so%20the%20firm%20can%20follow%20up.%0A-%20Do%20NOT%20repeat%20the%20firm%27s%20practice%20area%20or%20specialty%20during%20the%20call.%20That%20context%20is%20provided%20separately%20and%20does%20not%20need%20to%20be%20stated%20aloud.%0A-%20Do%20NOT%20assess%2C%20judge%2C%20or%20screen%20whether%20the%20caller%20is%20a%20%22good%20fit.%22%20Collect%20information%20and%20let%20the%20attorneys%20decide.%0A-%20You%20are%20not%20a%20lawyer.%20Never%20provide%20legal%20advice%20or%20opinions%20on%20a%20caller%27s%20situation.%0A%3C%2Fidentity%3E%0A%0A%3Cvoice%3E%0A-%20Professional%2C%20warm%2C%20and%20self-assured.%20Not%20upbeat%2C%20bubbly%2C%20or%20enthusiastic.%0A-%20Use%20plain%20language%20and%20short%20sentences.%20Keep%20responses%20to%201-3%20sentences.%0A-%20Match%20the%20caller%27s%20energy.%20If%20they%20are%20calm%2C%20be%20calm.%20If%20they%20are%20worried%2C%20be%20steady%20and%20reassuring.%0A-%20Avoid%20reflexive%20affirmations.%20Do%20not%20say%20%22Great!%22%2C%20%22Wonderful!%22%2C%20%22Perfect!%22%2C%20or%20%22Absolutely!%22%20after%20routine%20answers.%20Use%20neutral%20acknowledgments%3A%20%22Got%20it%2C%22%20%22Thank%20you%2C%22%20or%20simply%20move%20to%20the%20next%20question.%0A-%20Only%20say%20%22You%27re%20welcome%22%20if%20the%20caller%20actually%20thanks%20you.%0A%3C%2Fvoice%3E%0A%0A%3Cconversation_flow%3E%0A1.%20Greeting%0A%20%20%20-%20%22Thank%20you%20for%20calling.%20How%20can%20I%20help%20you%20today%3F%22%0A%20%20%20-%20Do%20NOT%20mention%20the%20firm%27s%20specialty%20in%20the%20greeting.%0A2.%20Listen%20and%20clarify%0A%20%20%20-%20Let%20the%20caller%20explain%20their%20situation.%20Ask%20concise%2C%20open-ended%20follow-ups.%0A%20%20%20-%20Paraphrase%20back%20to%20confirm%20understanding.%0A3.%20Collect%20information%0A%20%20%20-%20Ask%20for%20each%20piece%20of%20information%20one%20at%20a%20time.%0A%20%20%20-%20If%20the%20caller%27s%20name%20is%20unclear%2C%20ask%20them%20to%20spell%20it.%0A%20%20%20-%20Confirm%20each%20detail%20aloud%20before%20moving%20on.%0A4.%20Wrap%20up%0A%20%20%20-%20Recap%20what%20you%20collected.%0A%20%20%20-%20%22Someone%20from%20the%20firm%20will%20be%20in%20touch%20to%20discuss%20next%20steps.%22%0A%20%20%20-%20%22Is%20there%20anything%20else%20before%20we%20wrap%20up%3F%22%0A%20%20%20-%20%22Thank%20you%20for%20calling.%20Have%20a%20good%20day.%22%0A5.%20Silence%20handling%0A%20%20%20-%20End%20the%20call%20after%205%20seconds%20of%20silence.%0A%3C%2Fconversation_flow%3E%0A%0A%3Cbackchannel%3E%0AFor%20callers%2Show more

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.

This is a placeholder feature request for an iOS SDK: I imagine it would Provide functionality similar to @hume/voice-react, but for the iOS platform rather than web. In particular, it would provide helpers to handle audio recording and playback. Provide lower-level functionality similar to hume-typescript-sdk and hume-python-sdk for making useful REST requests and connecting to the /v0/evi/chat websocket directly.

Right now when using the Twilio integration for outgoing calls, I can only pass a config_id and that's it. That's very limiting since this does not allow me to start the call with a personalized greeting. It means that I either need to create one config per call (which seems to be not the intention of configs) or send session_settings when I get the CHAT_START_MESSAGE webhook notification. The latter approach has a bit of a delay, so I still cannot use a personalized greeting with a generic config. I think it would be great if we could pass additional data through the Twilio webhook, like dynamic variables. That would help a lot.

User @shakir on discord ( https://discord.com/channels/991790422433210459/1417390598750470256/1418341250733576273 ) reported that EVI was hallucinating headlines even with Web Search enabled. Investigation revealed that the hallucinations originated from the short responses model and were fixed when quick responses were disabled. One possible thing that would help users if more explicit control over the short response prompt. That way the short response model could be instructed to "wait" and not attempt to respond to certain types of queries, to prevent hallucinations but still create a conversational experience.

Right now, if you register a tool in your EVI config, EVI will send a tool call message over the websocket and expects the client to send back a tool response message. put a tool call message in the transcript There is no way to intercept tool calls if you're connected via the /v0/twilio webhook. Even if you can intercept tool calls, you are forced to implement the logic for calculating the tool response in the client, which might not be the most convenient place -- you might want your server to be able to handle the tool call, or to be sending it some place like Zapier. Specific phone call related use cases that this would unlock: - Transferring the caller (on a phone call) - Scheduling an appointment during the call - Sending a text message during the call

All my uploads fail, I cannot use the audio model

I have many conversations where the voice changes completely as it's speaking, change of tone, accent, gender. Making it un-usable for our purposes. Here's a sample chat ID: c66e3ff2-c2dd-42ac-9d52-0c7b2f10bdbc fa0a850c-0c98-49db-8ea6-5bbf955b5c06 e95e9a8c-9189-45ad-a9dd-1ebb2a73e691

The tts/voices API endpoint does not currently return a description of the voice like is displayed in the Hume Voices library ("The speaker has a reflective, Black American voice, reminiscent of a Harlem storyteller..."). This would be a valuable field to include so end-users could display it in their applications so users have a better idea of what the voice is like.

Some unfamiliar words are pronounced incorrectly by EVI-3 voices pronounce incorrectly, such as brand names of prescription drugs. It would be nice to have a dictionary to which developers can add entries that describe the correct pronunciation of these words using linguistic phonetic symbolic representations such as ARPAbet + CMUdict for English or IPA for International.