๐ One prompt, many languages, all on-device.
We are shipping multilingual support in React Native ExecuTorch v0.9:
โ Gemma 4 generates the response
โ Kokoro TTS speaks it aloud
โ Everything runs locally on mobile ๐ฑ
Here's a quick demo ๐
With RN ExecuTorch 0.9.0, you can run @OpenAI's privacy filter to redact sensitive data, on-device! ๐
What else is new?
โก๏ธPose estimation with YOLO โ detect and track body keypoints in real time
โก๏ธ FastSAM for promptable image segmentation โ tap an object, get a mask
โก๏ธ NPU-accelerated Whisper โ up to 10x faster speech recognition on iOS
Watch the privacy filter demo and check out the full release notes! โฌ๏ธ
@BoazWith Honestly the hardest part was picking which cat photo to use first ๐บ. Skia handles the cutout rendering, but the real magic is FastSAM doing instance segmentation fully on-device
I've always been a heavy user of iOS stickers - so I wondered how hard it'd be to build my own in React Native.
One hook, one model, a bit of Skia - that's the whole app and it's fully local too!
react-native-executorch 0.9 drops next week with FastSAM support ๐๐
Just got @OpenAI's privacy filter model running on-device with react-native-executorch.
~600MB RAM, and the quality is genuinely impressive.
Imagine scanning an email before you send it to catch sensitive data โ no cloud, no leaks. Tons of new on-device use cases unlocked.
Coming soon ๐
Listen, my friends - this is very important.
Right now in Poland we have the biggest charity action, helping young warriors fighting cancer. ลatwogang and Bedoes 2115 are doing a crazy good job for @fundacjacancer - already more than 38 million PLN raised. Big respect. But we can do more. Esports family - this is our moment.
Soโฆ I give something very special for donation. The most valuable skin from my collection โ iconic Karambit | Fade โmotar2kโ. For me? Priceless. So many memories, so many beautiful moments on my esports road. Tell meโฆ is there a more legendary knife in CS? Today, we make it even more special.
Simple rules - starting price $5,000. You write your offer in the comments under this post or DM me. Tomorrow at 1 PM I contact highest offer and we make everything through official charity channels.
Letโs show the world how strong the esports family is ๐ช
@realmotar2k@ohnePixel@Gaules@zipelCS@FalleNCS@Vitality_apEX@NiKoCS_@s1mpleO@ESL@BLASTPremier@pglesports@StarLadder_CS@HLTVorg@TeamFalconsGG@FaZeClan@TeamVitality
#cancerfighters #fundacjacancers
#ลatwogang #bedoes2115
@stevibe What does the memory utilisation look like? I wonder if it's possible to run the model on a MacBook and still have some wiggle room with compute to do other stuff.
OpenAIโs open-source model Privacy Filter just dropped ๐พ
- Automatically masks PII like phone numbers & addresses
- 1.5B parameter model which is mobile-capable
- Blur private data before it ever leaves your device
- Coming soon to React Native ExecuTorch ๐
spent some tokens vibe-coding a 3D embedding space visualization in React Native
the embeddings are computed locally using react-native-executorch, the visualization uses typegpu, co-authored by opus 4.7 max.
source ๐
You're in a jungle, no signal, and you just picked up something that's either a rare artifact or a really weird fruit ๐. What do you do?
You can take a photo and get an answer offline โ using Vision Language Models from the latest RN ExecuTorch release.
Check it out in our app Private Mind on the App Store and Google Play โฌ๏ธ
LFM 2.5 vl 450M uses less than 1GB of RAM and generates ~90 tok/s. Soon you will be able to run @liquidai models even on potato. Now available in react native executorch.
https://t.co/owvdRDqUJx
Qwen 3.5 from @Alibaba_Qwen just dropped - performance that was cloud-only a year ago, now running on your phone.
Multimodal LLMs are coming to react-native-executorch in the next release.
Here's a sneak peek of on-device image captioning ๐