Oti.

⚠️Security update on @pxxl_space . Since late October, we investigated a large abuse wave on Pxxl. Newly created Free-tier projects were used as outbound proxy/VPN relays, alongside remote-monitoring agents with command-execution capability inside affected containers. We treated it as high severity.

Today we are launching Vercel Security Review in public beta. Run vulnerability reviews on every push, and ship fixes in one click with Vercel Agent. Read the blog → vercel.com/blog/vercel-secur…

i built a free, local alternative to @superwhisper and @WisprFlow called Woof. it runs Bark 3.0 on device inspired by @0xSigil and team, tired of seeing people pay for something that takes a day to build super fast, free forever, secure, and damn adorable

Imagine spending 3 months building your mobile app, telling everyone it's launching on Friday, preparing your marketing, and getting your first customers excited. Then you try to publish it on Google Play and discover that you need 12 people to test your app for 14 consecutive days before you can even apply to launch publicly. 😭 And that's assuming all 12 testers stay opted in throughout the testing period. The funniest part is that after completing the 14 days, Google can STILL deny you production access if they're not satisfied with your testing. So you've built your app, paid your developers, prepared your launch campaign, and now Google is basically telling you to go and find 12 disciples. 😭😭 All because you opened a new personal developer account. I get why Google wants to reduce spam and low-quality apps, but imagine telling your investors your launch is delayed because Brother Tunde and 11 other people need to stay opted into your app's closed test for two weeks. 😭

They said African languages were too fragmented, too complex and too commercially insignificant to build advanced AI for. We disagreed. So we built it. 64 languages. 166 voices. 3 billion parameters. ONE model. Today, we are unveiling STORM-Ensemble 3.0, the most capable multilingual speech synthesis model we have built at PlotWeaver. From Yorùbá, Igbo, Hausa and Nigerian Pidgin to Swahili, Amharic, Wolof, Zulu, Xhosa and dozens more, STORM-Ensemble 3.0 generates speech across 64 languages, with 166 distinct speaker voices. Every voice you hear is generated by our model, not a prerecorded sample. But the real breakthrough isn't the numbers. It's what becomes possible when this technology is in the hands of people who can build with it. -African filmmakers distributing stories across languages. -Banks and telcos serving customers in their mother tongues. -Education platforms reaching more children. -Developers building voice AI products without starting from scratch. We are not building AI for African languages simply to prove it can be done. We are building the infrastructure to make our languages commercially viable at global scale. Next week, I will be speaking at the Commonwealth Business Forum in Antigua about unlocking the Commonwealth's creative potential. This launch is a practical demonstration of that ambition. Africa must not simply consume the future of AI. We must build, own and shape the infrastructure that powers it. Watch the video. Listen to the voices. Imagine what's possible. Introducing STORM-Ensemble 3.0. PlotWeaver. Build Once. Speak Every Language. #PlotWeaver #STORMEnsemble #VoiceAI #AfricanLanguages

Millienials learned how to do it best by creating systems

Macro average WER 7.71%

Today we’re introducing Reasoning for Realtime TTS-2, making it the first TTS model in the world that can steer itself based on context. In consumer apps, your LLM generates every reply. Directing how each one sounds takes extra prompts, tags, and rules. Reasoning now handles that inside your TTS request. It reads the text and the context you pass, writes a delivery direction, which Realtime TTS-2 performs. For developers this means: - Emotion, tone, and pacing shaped by the context of each line. - No emotion-tag rules to maintain in your chat prompt. - Fewer LLM output tokens spent writing voice directions. The generated direction comes back in the API response, so you can inspect it. Your own instructions and inline tags always take precedence. Available today with Realtime TTS-2. Who are you building for?

Yes. RealtimeSTT has a ready omnilingual_asr engine (supports omniASR_CTC_300M_v2 and larger) with built-in VAD and live transcription—just point it at the model and drop into Python. omniASR-server gives an OpenAI-compatible REST + WebSocket API so you can stream audio chunks and get text back with almost no custom code. Both abstract the VAD + CTC pipeline. For pure offline/low-resource, sherpa-onnx + the int8 Omni CTC models also works cleanly.

Jesus lived 600 years earlier in a far more unrefined era, yet He never owned a single slave. What is most startling about Christ is how He entirely bypassed the cultural traps of antiquity. He easily could have engaged in the socially permissible norms of His day that age horribly now; like marrying a minor, or owning and sexually exploiting slaves. But He didn’t. He lived with absolute moral perfection in a time when the ethical bar was on the floor, and His standard remains flawless today. He is God.