This morning, I received a email from OpenAI saying that they launched a new speech-to-speech model. I was surprised to find that the default model does not point to the latest gpt-4o-realtime-preview-2025-06-03; I remember that would automatically point to the latest version.⁉️

Sokuji now on Edge! 🎉 Our real-time voice-to-voice translator for online meetings is now available as an Edge extension. Break down language barriers on your favorite browser! 🔗 microsoftedge.microsoft.com/addons/detail/…

除了Claude感觉没一个会写代码的 （难道是只有Claude能改自己写的代码？ 睡觉去了😤

PipeWire徹底解説：LinuxマルチメディアアーキテクチャをOSエンジニアリングの観点から読み解く qiita.com/shinonome_taku… #Qiita via nano

日本のソフトウェアエンジニアにとってのAIツールの影響、企業方針、副業、コストのトレードオフ qiita.com/shinonome_taku… #Qiita via nano

We just added support for Palabra.ai in Sokuji — our 3rd AI provider after OpenAI & Gemini! Tested and it works amazingly well — plus, free users get 30 mins/month 🆓 Still pending Chrome Web Store approval... But I can't wait — grab it now from GitHub: github.com/kizuna-ai-lab/…

Mu Li was the prof who first got me into Transformers/GPT. Now he’s launching multimodal models—today they open‑sourced Higgs Audio v2 (10M+ hrs pretrain, super expressive TTS). Demo + code: youtube.com/watch?v=0Dv4s2… github.com/boson-ai/higgs… Sokuji will add voice cloning soon.

こんなんじゃ、知らない人に dzmm.ai 使ってると勘違いされるって〜

🚀 Want to use real-time AI translation in your Zoom meetings? Check out this step-by-step guide for setting up Sokuji with Zoom! It’s simple, open source, and supports GPT-4, Gemini, and more. 🌐🎙️ 👉 kizuna-ai-lab.github.io/sokuji/tutoria… #AI #Zoom #Translation #OpenSource #Sokuji

I’m building a product with end-user usage-based billing (e.g. per minute/token/request) But most platforms like Clerk or RevenueCat don’t support this model directly Any recommendations for services or tools that make metered billing easier? Would love to hear what others using!

I have a small hand, I love small smart phone.

記事を投稿しました！ 2025年のDuckDB on Lambda：サーバーレス分析の革命 [AWS] on #Qiita DuckDB #AWS #typescript qiita.com/shinonome_taku…

community.cloudflare.com/t/byok-configu… We have discovered that BYOK configuration is not being applied when connecting to OpenAI’s Realtime API via WebSocket through Cloudflare AI Gateway. Despite having BYOK properly configured and Authenticated Gateway enabled.

BYOK works normally with interfaces using the HTTP protocol, but does not work properly with interfaces using the WebSocket protocol. Cloudflare Developers Cloudflare

It looks just like a programmer who has just learned React, with useEffect everywhere.