Sesame, starting behind the viral virtual assistant Maya, releases its base AI model
You have a company Sesame released the base model that feeds yeast, impressively realistic voice assistantS
The model, which is 1 billion parameters (“parameters” relating to individual components of the model), is below the Apache 2.0 license, which means that it can be used on a commercial network with few restrictions. Called CSM-1B, the model generates “RVQ audio codes” from text and audio inputs according to Sesame description of the AI ​​Dev platform hugging the faceS
RVQ refers to “residual vector quantization”, an audio encoding technique in discreet markers called codes. RVQ is used In a number of recent AI audio technologyIncluding Google and Encodec’s Soundstream of Meta.
CSM-1B uses a model of Meta’s Llama family As his spine is paired with the audio component “decoder”. A finely tuned CSM Powers Maya version, Sesame says.
“An open source model here is a basic generation model,” Sesame wrote in CSM-1B’s Hug and Girub storage. “It is capable of producing different votes, but it is not refined in any specific voice (…) The model has some capacity for non -English language due to pollution of data in learning data, but it probably won’t do well.”
It is not clear what kind of data sesame he used to train CSM-1B. The company did not say.
The model has no real precautions to talk about, it is worth noting. This is a situation of the “honest system”. Sesame simply urges developers and users not to use the model to mimic a person’s voice without their consent, to create misleading content as false news, or to participate in “harmful” or “malicious” activities.
I tried demonstration It took me less than a minute on the hug and cloning my voice. From there, it was easy to generate a speech of the desire of my heart, including on controversial topics such as elections and Russian propaganda:
Sesame, co-founder of the co-creator of Oculus Brendan Iribe, became a viral at the end of February for his assistant technology that approaches the clearing of the UNCANNY VALLEY territory. The other assistant of yeast and sesame, miles, take in breaths and talk with differences and can be interrupted as you speak, Like the Openai voice modeS
Sesame raised an undisclosed amount of capital from Andreessen Horowitz, Spark Capital and Matrix Partners. In addition to the construction of a voice assistant technology, the company says that the AI ​​glasses are “designed to wear throughout the day”, which will be equipped with their personalized models.
English