rohan

launching silk mulberry 1.5 one of the fastest multilingual voice models in the world it matches the best voice models in quality benchmarks (MOS) all this at more than 95% lower cost ₹0.40/min (~$0.0046/min) try now 👇

intelligence has always been scarce. and anything scarce gets metered. doctors, engineers, lawyers were always gated, to get access you needed to pay by hour/ project. ai was supposed to break this meter. but it didn't. it costs billions to train and infer the models, and that bill lands on you in form of a meter sliced into tokens and limits. remember kodak? our parents used to click pictures with films - and because each film had a cost, you had to wait for the perfect pose, only then would you click. but as soon as camera became digital, photos became abundant. now for the same pose you will click 50 photos, but this was considered waste, why are you wasting. but this waste led to a new habit, and this new habit led to instagram and snapchat. the entire foundation of these products is on the principle of photos becoming too cheap to meter. now what does too cheap to meter mean? it doesn't mean free, but it means a point....... where cost becomes something that i don't worry about. it's the moment where the technology becomes cheaper than counting. we start to think of it as abundant, and at this point something changes entirely -waste is a signal. the moment people start wasting a thing is the moment a new behavior is born. ai is still in the meter era. look at how we use ai right now, you hit a limit and wait till tomorrow. you start optimizing your questions, you show your prompt engineering skills, you learn the magic of context harnessing - but all of this is just a simple sign that intelligence is still scarce and expensive. we are living in the meter era of ai. the day intelligence becomes abundant, we'll forget that we ever did this, the same way no gen Z knows that films ever existed. and this is the sole reason why we haven't seen any consumer ai product achieve a breakthrough. we check instagram 90 times a day, that is a habit. and no one has built the instagram of the ai era, because the meter won't let them. rumik is built on the idea of making frontier intelligence cheap. we started with the most expensive modality - voice. just launched silk 1 - our voice models that match the frontier in quality at 95% cheaper costing less than half a cent per minute. and remember what i said. abundance will show up first as waste and people will laugh over it and taunt - why are you wasting intelligence. and the first signs are here. as soon as we launched our api's, one person is building an app, where you can point to any thing - your chair, fridge, painting - and you can start talking to it. a talking fridge is useless and that's the point. it looks as stupid as taking 50 photos for one good pose in 2005. the useless thing is always the first thing. there's no word for this yet. there wasn't a word for selfie either, until photos became too cheap to meter. the behavior always arrives before the product. i can see the behavior is here, the word and next set of products will follow.

ira just got an upgrade canvas that knows what time it is. added tic tac toe. good luck. talk to ira in voice while you play yt shorts, chess, or tic tac toe. and a more thoughtful design. dropping soon...

jevons paradox is coming for intelligence!

everyone is shipping voice this week. listen to ours.

meet silk mulberry 1.5, a voice model built to outperform expectations. - 162ms end-to-end latency - replies before you finish reading the message. - matches elevenlabs v3 on the MOS benchmark while being 95% cheaper. (₹0.40/min vs ₹9.40/min) ⁃ you don't pick a voice, you design one - describe the age, accent, pitch, emotion, whatever you want, only limit is your imagination. - mid-sentence language switching across 20+ languages with native accents, plus expressive range (whispers, singing and much more)

i am building a new growth team at @rumik_ai two role i'm hiring for a) growth lead b) growth intern (3X) what i need from you - growth lead 1. managed multi-crore performance budgets 2. you can drill down to understand ad level retention 3. can write a 4 page article in normal english in 30 mins, without using AI 4. have impeccable taste and extremely high agency growth intern 1. experienced this role before 2. publish 5 content everyday 3. have impeccable taste and extremely high agency our interview process includes a very deep 3-4 day thorough assignment (so if you don't like putting in the work - skip it) to apply - DM me or email at hiring@rumik.ai

if you're a founder or CEO - the single biggest advice i would give you is FIRE YOUR SENIOR EMPLOYEES - they don't get AI, and only 0.1% people are even trying to go in depth HIRE PEOPLE COMING OUT OF COLLEGE 365 days later, you'll thank me

bangalore is the only place on one side you'll find a pine tree and on the other a coconut tree

@hm how do i return damage article. I have been finding ways on your app there is no option

@infinitipill @rumik_ai 🙏

@rumik_ai crossed 1 million+api request. LFG🇮🇳🥳 Top 3 models in the world 20x cheaper. Try rumik.ai

silk by @rumik_ai (our voice model) has now reached 1 million+ api requests in 10 days of launch we have gone from 100k to 1mn in 5 days super grateful to all the companies & builders building on top of silk on benchmarks - silk is top 3 models in the world at only 0.4rs/min (20x lower cost) try yourself, link in comments if you are trying to move your workloads to silk, we will help you with the setup (just DM/ email me)

hiring for two roles at @rumik_ai 1. growth intern (2X) skills you need to have - - script writing, shooting and video editing - push out 2 creatives/day - ai-generated films pay - 50K/month in person, delhi chef prepared lunch, dinner everyday :) full time perks are insane, and listed on our careers page to apply -> DM/email me (rohan@rumik.ai) with your best work

@InsaaniManav @rumik_ai hey we work all 7 days

@rumik_ai @lets_dig_deeper I'm looking to start a builders community in Delhi . Invite only . Working professionals . Looking for a venue would you folks be open to host ? DMs open for anyone else interested

@Jay_dobariya_ write in latin script

@lets_dig_deeper Try gujarati in playground, not working, is that anything i need to do for getting gujarati result ?

launching silk mulberry 1.5 one of the fastest multilingual voice models in the world it matches the best voice models in quality benchmarks (MOS) all this at more than 95% lower cost ₹0.40/min (~$0.0046/min) try now 👇

@rumik_ai try now - rumik.ai

we evaluated mulberry on instructTTS eval, a benchmark built to test how well TTS systems follow the given voice description, eg: "sound excited" or "speak slower" etc. the full benchmark results are in the table below.

@PriGoistic @rumik_ai yes

@lets_dig_deeper @rumik_ai do u guys have research roles open ?

hiring a growth lead at @rumik_ai looking for someone with deep experience and understanding on performance marketing things that are must 1. you have managed multiple crores in ad budgets 2. you are great at maths 3. you can analyse billion rows of data, make cohorts and gather insights 3. can write a 4 page article in normal english in 30 mins, without using AI 4. you have impeccable taste and extremely high agency our interview process includes a very deep 3-4 day thorough assignment (so if you don't like putting in the work - skip it) _ for us, the team is the most important part of the company and that's why - we pay more than anyone else. (not because we have to. because we want the best people and we respect what they're worth) and beyond salary: a) chef prepared breakfast, lunch and dinner b) health insurance for you + family c) quarterly offsites - goa, bali, phuket etc d) high ESOP's _ if this excites you - write an email to hiring@rumik.ai