Topics
later
AI
Amazon
Image Credits:Bryce Durbin
Apps
Biotech & Health
Climate
Image Credits:Bryce Durbin
Cloud Computing
commercialism
Crypto
go-ahead
EVs
Fintech
Fundraising
Gadgets
Gaming
Government & Policy
ironware
Layoffs
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
Security
societal
outer space
Startups
TikTok
Transportation
speculation
More from TechCrunch
result
Startup Battlefield
StrictlyVC
newssheet
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
AI companySesamehas released the base exemplar that power Maya , theimpressively realistic articulation assistant .
The role model , which is 1 billion parameters in size ( “ parameter ” referring to individual components of the model ) , is under an Apache 2.0 licence , entail it can be used commercially with few restrictions . Called CSM-1B , the model generates “ RVQ audio frequency code ” from text and audio input , according toSesame ’s description on the AI dev platform Hugging Face .
RVQ refer to “ residual transmitter quantisation , ” a technique for encode audio into distinct token called code . RVQ is usedin a figure of recent AI sound recording applied science , include Google ’s SoundStream and Meta ’s Encodec .
CSM-1B use a model fromMeta ’s Llama familyas its gumption paired with an audio “ decipherer ” element . A fine - tuned variant of CSM big businessman Maya , Sesame says .
“ The model open - sourced here is a base coevals example , ” Sesame writes in CSM-1B’sHugging FaceandGitHubrepositories . “ It is capable of producing a variety of spokesperson , but it has not been fine - tuned on any specific spokesperson [ … ] The example has some capacity for non - English languages due to data pollution in the grooming data , but it in all likelihood wo n’t do well . ”
It ’s unclear what data point Sesame used to train CSM-1B. The companionship did n’t say .
It ’s worth noting the model has no real safeguards to speak of . Sesame has an honor system and but recommend developers and user not to use the mannikin to mime a person ’s voice without their consent , create misleading content like phoney news , or lock in “ harmful ” or “ malicious ” activities .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
I triedthe demoon Hugging Face , and cloning my representative took less than a minute . From there , it was easy to generate speech to my substance ’s desire , including on controversial subject like the election and Russian propaganda .
Consumer Reports recently warned that many democratic AI - powered voice cloning tools on the marketdon’t have “ meaningful ” safeguardsto prevent fraud or abuse .
Sesame , co - founded by Oculus Colorado - Lord Brendan Iribe , went viral in belated February for its assistant tech , which comes close to clearing unearthly vale dominion . Maya and Sesame ’s other assistant , Miles , take breaths and address with disfluencies , and can be interrupted while address , much like OpenAI ’s Voice Mode .
Sesame has raised an undisclosed amount of capital from Andreessen Horowitz , Spark Capital , and Matrix Partners . In addition to building voice assistant tech , the company say it ’s prototyping AI deoxyephedrine “ plan to be wear out all day ” that ’ll be fit with its tradition models .
