←
Ettore Di Giacinto3,382 posts
Ettore Di Giacinto avatar

Ettore Di Giacinto

@mudler_it

dad, creator of LocalAI( @LocalAI_API localai.io) and Kairos (kairos.io) , ex @SUSE/@Rancher, ex-Gentoo Dev. vllm.cpp and APEX quants

◷ Joined January 2016▧ 291 Photos and videos
Ettore Di Giacinto avatar
Ettore Di Giacinto@mudler_it·Oct 3

Your local agent can now hear what was said, who said it, and what happened around them. parakeet.cpp transcribes audio, separates speakers, detects 527 sound classes, and now enrolls and remembers named voices across recordings. One pass, local C++.

0:55
Ettore Di Giacinto avatar
Ettore Di Giacinto@mudler_it·3h

Yes! and.. it's coming :)

Found some time to play a bit more with it and holy shit, I love it! Just using now the built in functions, which does work fine, but the interface rework, ability to stop and inject new prompts and the compaction feature - works like a charm!

Ettore Di Giacinto avatar
Ettore Di Giacinto@mudler_it·Nov 13

I've been quiet a bit, but been really focused on @LocalAI_API . Next release is gonna be really nice. I'm so excited for it. If you point your harness to LocalAI's master now it's gonna go crazy for the things it can do with it.

FluidVoice avatar
⟳ Ettore Di Giacinto retweeted
FluidVoice@FluidVoiceApp·Nov 13

Today we’re releasing Fluid Speech Pro : #1 open ASR model under 1B #2 on @huggingface OpenASR leaderboard beating models up to 50x its size. The current SOTA models are huge, closed or paid, so we decided to make our own. Supports 25 Languages and custom built to run on your everyday devices including your phones. it's free and open weights on HF Today!

Post attachmentPost attachmentPost attachment
Niels Gron avatar
⟳ Ettore Di Giacinto retweeted
Niels Gron@niels_gron·Nov 4

I've worked at software startups for 22 years, including my own, and I've invested in startups for 10 years. None of them would have been possible without OSS. I am trying to give back to the community. In addition to supporting @ashxhart, I've also provided hardware to @JoshuaSWarren, @mudler_it, and @antirez for their Local AI work. I believe their work will open up use-cases that future startups will be built on, and builders like them deserve way more recognition. I will try to pay things forward as much as I can. If there are any successful founders out there who have benefited from OSS, pitch in—the future looks great.

The community hasn’t finished with me yet 🤯🤯 Honestly, I’m speechless at the generosity and kindness shown by everyone. @niels_gron reached out to me directly and donated a serious amount towards TensorFold which has enabled me to purchase even more AI hardware. Unbelievable 🫶🏼 If you have an OOS project that people depend on then comment below 👇🏼

Ettore Di Giacinto avatar
Ettore Di Giacinto@mudler_it·Nov 3

With all the respect, no, this is not the right attitude. Some trade speed for quality, some don't. Seeing demo quality is ok, and makes sense for experimentation, but hey, not gonna replace my llama.cpp with a low quant that can't say a single word in my language without making mistakes. Not sure for what should I use it for?

Watched the latest @LinusTech video and they are running QWEN 3.8 27B with pipeline parallelism GGUF on LM studio Q4KM and complaining it isn’t fast. Oh my eyes. Please PLEASE someone who’s close to him educate him on VLLM and Strata. He could literally be getting 5x better speeds and better intelligence. Sure the setup is complex but it’s a one time setup with a cloud agent.

Ettore Di Giacinto avatar
Ettore Di Giacinto@mudler_it·Oct 30

Agree 💯. What I tend to build instead mostly maintenance factories, which is the boring part usually, which drains time even for simple tasks.

a "software factory" might be the thing i least desire to use LLMs for. i want a semantic workbench for supercharging artisanal software production. i want maximum leverage, not maximal throughput.