正在加载视频...

视频加载失败

Introducing Saaras V4 Multi-speaker Our best-in-class speech recognition model that accurately captures overlapping conversations and multi-speaker interactions, while delivering state-of-the-art performance in English across accents.

31,374 次观看 • 2 个月前 •via X (Twitter)

28 条评论

Abhay kumar 的头像
Abhay kumar2 个月前

sarvam ai admin carefully finding the perfect indian news debate clip to demonstrate saaras .

nj 的头像
nj2 个月前

Thank you ye bahaut jaroori tha. Now I hope TV channels live individual mute ka option le aaye 😂

Anant Jain 的头像
Anant Jain2 个月前

thank you @SarvamAI the day is not far, when I am able to watch all my favourite Japanese anime in Hindi, with incredible accent with emotion and thrill'😍

Tpot 的头像
Tpot2 个月前

finally an ASR model trained for the hardest benchmark: 5 founders talking over each other at a tpot dinner

सूर्यः 的头像
सूर्यः1 个月前

Is this feature is available in Indus app

सूर्यः 的头像
सूर्यः1 个月前

- Please add an Edit Message feature. - Copy and Select All options are not working properly for prompts. - Please make the Enter key create a new line instead of sending the message, or add an option to choose this behavior. Thanks for continuing improve in Savram Ai model

Shraajan 的头像
Shraajan2 个月前

insane example to choose bruv

Observation Post 的头像
Observation Post2 个月前

@SarvamAI has taken it upon themselves to take india on the high pedestal of AI leaders and seem to be on the right path too. On the sidenote- they should have chosen a voice clipping of the Arnab Goswami show to really test this voice model

Adithya Gounder 的头像
Adithya Gounder2 个月前

When asking #SarvamAI questions via voice, if I pause the response and try to listen to it again, it restarts from the beginning; this needs to be fixed. @SarvamAI @SarvamForDevs @pratykumar

Bharatiya 的头像
Bharatiya2 个月前

Keep building India

Foxquart 的头像
Foxquart1 个月前

incredible product surely will help in making things more easy!

Neeraj Kumar 的头像
Neeraj Kumar1 个月前

Overlapping speakers are where speech recognition becomes genuinely difficult. Accuracy by accent, speaker count and overlap duration would make this especially valuable for meetings, support calls and public service applications.

Jaya Nayak 的头像
Jaya Nayak2 个月前

Accurately parsing overlapping conversations and multiple speakers is one of the hardest roadblocks in audio processing.🤩

Youth 的头像
Youth2 个月前

saaras v4 sounds impressive but can it handle non-english languages and dialects as well or it's still a work in progress?

Sushant 的头像
Sushant2 个月前

Applied for a job at @SarvamAI but never got any response. I am willing to relocate to India and contribute to building our own AI. I bring in lots of infrastructure and software engineering experience. Its just not clear how you are hiring at the moment.

Alysium Corp Studios Inc 的头像
Alysium Corp Studios Inc2 个月前

Is this available in South Africa?🇿🇦

cornball.dev 🐳 的头像
cornball.dev 🐳2 个月前

😅

Arghya 的头像
Arghya2 个月前

cant wait to use this

John Paul Reddy 的头像
John Paul Reddy1 个月前

one use case might be to understand the tv debates.

गौरव Gaurav 的头像
गौरव Gaurav2 个月前

Your auth is broken!

Yash Rastogi 的头像
Yash Rastogi2 个月前

😂😂

Sk Samim 的头像
Sk Samim2 个月前

whats the practical use case of it?

krayX 的头像
krayX2 个月前

Amazing

Madhumitha Kolkar 的头像
Madhumitha Kolkar1 个月前

I was mind-blown by the products that were showcased at Epoch yesterday ! Especially Bulbul V4 and Kivi ! Great work @SarvamAI !!

ᱫᱩᱞᱟᱹᱲ 的头像
ᱫᱩᱞᱟᱹᱲ2 个月前

when can we use it?

nkcbuilds 的头像
nkcbuilds2 个月前

now this is nice

arnab kumar 的头像
arnab kumar2 个月前

Sir ,give me download link of sarvam ai in gguf format ,only 5 gb in size for running on mobile.

Rhythmic Analyst 的头像
Rhythmic Analyst2 个月前

Wow!

相关视频