Diwas retweeted
nepali ai would move a lot faster if the government ran on the same internet as us.
what builders here need:
gpus we can pay for in rupees
government data, opened up
rules written for ai, not around it
anyways, we're building swar. open nepali voice cloning, try for free : naamchelabs.com/playground
when's the meeting? 🇳🇵
के तपाइलाइ थाहा छ नेपालमा गेम बनाएर संसारभर नाम कमाएका गेम कम्पनीहरू पनि छन । संसारकै सबैभन्दा चर्चित लुडो (~ २०० मिलियन Downloads ) र कलब्रेक गेम (१५० मिलियन + downloads) नेपालीले नै बनाएका हुन् । हामीले वहाँहरु लाई बोलाएर समस्याहरू पनि बुझ्यौं र यो Game Development क्षेत्रलाई सजिलो बनाउन र उद्योगको रूपमा अघि बढाउने कुराहरू बारे पनि कुरा गर्यौं र अब वहाँहरुको समस्या समाधान गर्ने गृहकार्य अघि बढाइएको छ । नेपाल बाटै हामीले Global Unicorn कम्पनी बनाउन सहज पार्नुपर्ने नियम, वातावरण सबै पुनरावलोकन गर्नेछौ ।
IT को प्रोडक्टहरु र एक्सपोर्ट पनि अहिले निकै promising स्टेजमा छ र हाम्रो ट्यालेन्ट, कम्पनीहरूले सहजै कमाउने, re invest गर्ने कुराहरू लाई ध्यानमा राख्दै tax सहजता र Policy improvement, international income मा conservative नियमहरू अझै सहज बनाउने छौं।
यसका साथै Game Jam, Startup events, IT Sector र राष्ट्र बैंक, अर्थ मन्त्रालयको regular interaction, Hackathons र समग्र IT Product & Gaming industry लाई फाईदा हुने ecosystem events मा invest गर्ने छौं ।
Diwas retweeted
yesterday we launched स्वर 🇳🇵
35k+ of you watched it, and so many of you asked "how do i try it?"
so today we're introducing the स्वर playground 🎙️
→ 9 ready voices, or clone any voice from a 20 to 30 second voice note
→ type nepali, english, or code-switch mid-sentence
→ every line makes a few takes and keeps the clearest one
→ play, download, and replay everything from your history
→ free to try in your browser
go clone your voice and hear it speak nepali : )
naamchelabs.com/playground
presenting swar स्वर 🇳🇵
nepali has about 32 million speakers, yet almost no open text-to-speech sounds like someone from nepal.
we built swar, an open nepali voice model. give it 20 to 30 seconds of anyone's voice & it speaks for them: nepali, english, and everything in between.
on NepTTS-Bench, swar has the lowest pronunciation error: 2.92% beating omnivoice at 4.32%, elevenlabs v3 is at 12.91%.
it's #1 on the nepali voice cloning benchmark (fleurs, clean track), and makes half the errors on nepali mixed with english: 5.6 vs 11.4.
the code, weights, benchmark and a free colab notebook are all open. runs on your own gpu.
only clone a voice with that person's permission.
github: github.com/Naamche-Labs/swar
hugging face: huggingface.co/Aarjan/swar
website: naamchelabs.com/models/swar
Diwas retweeted
yesterday we launched स्वर 🇳🇵
35k+ of you watched it, and so many of you DM'd asking "how do i try it?"
so today we're introducing the स्वर playground 🎙️
→ all nine voices, ready to play
→ nepali, english, or both in one line
→ clone your own voice from a 30 second voice note
→ free to try
go clone your voice and hear it speak nepali : )
naamchelabs.com/playground
Diwas retweeted
presenting swar स्वर 🇳🇵
nepali has about 32 million speakers, yet almost no open text-to-speech sounds like someone from nepal.
we built swar, an open nepali voice model. give it 20 to 30 seconds of anyone's voice & it speaks for them: nepali, english, and everything in between.
on NepTTS-Bench, swar has the lowest pronunciation error: 2.92% beating omnivoice at 4.32%, elevenlabs v3 is at 12.91%.
it's #1 on the nepali voice cloning benchmark (fleurs, clean track), and makes half the errors on nepali mixed with english: 5.6 vs 11.4.
the code, weights, benchmark and a free colab notebook are all open. runs on your own gpu.
only clone a voice with that person's permission.
github: github.com/Naamche-Labs/swar
hugging face: huggingface.co/Aarjan/swar
website: naamchelabs.com/models/swar
Diwas retweeted
Sick Slider Effect in AE With Simple Expressions
1. Create your shape layers.
2. Create a Null, parent all shape layers to it, and name it Controller Null.
3. Create another Null, parent Controller Null to it, and name it Center Null.
4. Select the top shape layer, go to Scale, and paste:
center = thisComp.layer("Center Null").transform.position;
itemPos = thisLayer.toComp(thisLayer.anchorPoint);
d = Math.abs(center[1] - itemPos[1]);
maxDist = 300;
minScale = 70;
maxScale = 100;
s = ease(d, 0, maxDist, maxScale, minScale);
[s, s]
5. Go to Opacity and paste:
center = thisComp.layer("Center Null").transform.position;
itemPos = thisLayer.toComp(thisLayer.anchorPoint);
d = Math.abs(center[1] - itemPos[1]);
maxDist = 400;
minOpacity = 20;
maxOpacity = 100;
ease(d, 0, maxDist, maxOpacity, minOpacity)
6. Copy both expressions and paste them onto all your shape layers.
7. Done!
Diwas retweeted
HIRING
Looking for skilled video editors who can edit in this style.
If you think you can do it, drop your work in the comments below.
Video credit: @atomikgrowth
Diwas retweeted
Out of 500+ product launches on X this year, a big share came from products with relatively small audiences.
Some still managed to generate 500k+ views on launch day.
But views alone don’t matter if they don’t turn into actual users.
The launches that performed best usually had 5 things in common:
1. Easy to understand
People could figure out what the product did almost instantly.
2.Easy to try
No waitlist, no sales call, no unnecessary friction. People could get into the product within a minute.
3. The product created proof
Screenshots, outputs, demos, results, templates and before-and-afters gave people something worth sharing.
4.Real usage happened early
Not just likes and impressions. Actual users started creating feedback, screenshots and social proof in the first few hours.
5.They knew the baseline
They already knew their normal signup numbers, so they could clearly see what changed after launch.
You can have the best product in the world.
If people can’t understand it, try it and get value from it quickly, the launch won’t matter.
This is the launch we did for @kasada_io
Probably one of the best intros we've edited for @Laraacostar





