Community

Hi! I had reached out a few weeks ago to [[email protected]](mailto:[email protected]) to ask about the cost of gaining access to Hume's datasets for commercial use. I was wondering what that cost would be? Would it be possible to chat with someone further on this? Thank you!

Do you have a demo of the experience and deliverables?

\[{"source": {"url": "****************\*\*\*\*****************"type": "url"}, "results": {"predictions": \[{"file": "4599190.mp3", "file_type": "audio", "models": {"prosody": {"metadata": {"confidence": 0.9941406, "detected_language": "en"}, "grouped_predictions": \[]}, "language": {"metadata": {"confidence": 0.9941406, "detected_language": "en"}, "grouped_predictions": \[]}}}], "errors": [{"file": "4599190.mp3", "model": "prosody", "message": "unexpected error analyzing media: CUDA error: misaligned address\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\n"}, {"file": "4599190.mp3", "model": "language", "message": "unexpected error analyzing media: CUDA error: \n"}]}}]misaligned address\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

Hi anyone implement the APIs in Flutterflow? I looking for the text evaluation API.

I'm struggling right now to get this working. I have an audio stream coming in from TwilioFlex that I'm successfully calling a different web socket client with so pretty sure the issue is with how I'm trying to call Hume. Any help greatly and gratefully appreciated!

"message": "unexpected error analyzing media: CUDA error: an illegal memory access was encountered\\nCompile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.\\n"

Will using the Language model with a text file (with 100% correct transcription) provide worse results than an audio file with 90% correct transcription. Basically does the lang model only take into account text or does it take into account more complex ideas aswell.

Hi, I tried joining the Discord but the invite was invalid. Do I have to use my hume dev email to signup or something? Further, I used the batch API to generate sentiments and after labels titled: Sentiment 1, 2...etc...what do they represent? Lastly, how do I get Hume to spit out an overall sentiment score for a piece of text/audio/or video?

Eye contact is extremely important in the domain i am working in.... Is there a way to tell using Hume if someone is maintaining eye contact?

At this time we do not support emojis.