whispering about a secret tts

my head’s buzzing with code and possibilities. rn i’m splitting my time between my sde job, uni, and what nexora(or something else) might be —though i’m not sold on the name yet. work’s been intense, and i’ve also got this side project i’ve been keeping quiet about. figured it’s time to share a bit, so here’s what’s up. the deep research engine we're building at xAGI is making tremendous progress. it’s ai that chews through wild amounts of data—pitch decks, market reports, founder profiles—to give vcs the kind of insights that make or break investments. not just search; it’s like handing them a crystal ball. i’ve been wrestling with ml models that need to handle these messy, unstructured data, like half-baked funding docs or random linkedin posts. a few nights ago, i was up late fixing a python script that was tripping over malformed csvs—funding rounds with missing dates or weird symbols. i ended up hacking together a preprocessing pipeline with some regex tricks to clean it up. today, we got the RAG pipeline workign on it in addition to the Normal and DeepSearch modes. that brings me to this tts model i’ve been tinkering with on the side. i haven’t told many people, but i built a text-to-speech system that’s acctually something im proud of—it spits out human voices so real you’d swear it’s your friend talking, not a machine. the great part is: it’s stupidly efficient, running on way less compute than most models out there. i’ve been testing it on my linux box, feeding it random texts and hearing it nail accents, emotions, even little vocal quirks. i trained it with some ml tricks i picked up, keeping the model lean so it doesn’t need a supercomputer, infact, i was able to train it on just normal specs. i’m not ready to spill all the details—still figuring out what to do with it—but i’m itching to take it somewhere. maybe a startup, maybe open-source it, maybe plug it into something bigger. im super excited and have got alot in mind. but im sticking to figuring things out for now. maybe this might comeout as a surprise. all this is making me rethink nexora. i’ve been saying it’s about connecting startups—founders, coders, investors—but i’m wondering if i’m missing the mark. i need to formulate a good plan. i’m not sure if it’s a platform, an api, or something else, or if nexora’s even the right name. but i want this next thing to hit harder. the tts feels like a piece of the puzzle, but i don’t know where it fits yet. i’m cool with the uncertainty—it’s like coding without docs, just you and the problem. i’ll keep poking at it, see where it leads. — ciao, basab