Oruk (@OrukLabs) header
Oruk (@OrukLabs) profile picture

Humanizing Speech AI. Backed by a16z @speedrun.

San Francisco, CA
Joined April 2026
Photo by Oruk (@OrukLabs)Photo by Oruk (@OrukLabs)Photo by Oruk (@OrukLabs)Photo by Oruk (@OrukLabs)Photo by Oruk (@OrukLabs)Photo by Oruk (@OrukLabs)Photo by Oruk (@OrukLabs)Photo by Oruk (@OrukLabs)Photo by Oruk (@OrukLabs)Photo by Oruk (@OrukLabs)
Pinned Tweet
TONE. OF. VOICE. MATTERS.
Voice AI is getting the words right, but it still doesn't understand tone. We paused our PhDs at Stanford to build @OrukLabs, backed by a16z @speedrun. We train speech models to understand what you say AND how you say it. DM me for preview access to our most powerful model yet
2
11
3,074
We’re proud to sponsor EMNLP 2026! @emnlpmeeting At Oruk, we’re building AI that understands not just what people say, but how they say it. Excited to support the community advancing language and speech research. Humanizing speech AI. #EMNLP2026
🤖 Made with AI
3
2
1
9
866
This is local time reversal: sound runs backward inside each slice, but the slices stay in order. Short slices can leave speech understandable. Longer ones disrupt more of the cues. There's no single cutoff for every listener.
4
3
124
39,048
Speech can survive being flipped backward every twenty milliseconds. The same speech in backward pieces of 200 then 100 then 50 then 20 milliseconds. The original comes last. When do the words become clear? A listening demo from @OrukLabs.
89
144
60
2,339
637,095
"I've got him" is a bold claim from a human sled. She's being dragged through the mud by a giant dog: "It's fine! I've got him!" The subtext: "I am a sled now." Words say confidence. Delivery tells another story.
1
1
2
480
some workplace complaints begin with "smells amazing." Her coworker is microwaving an entire salmon. She says: "Mmm. Smells amazing." She means: "I'm calling HR." The transcript is missing the complaint.
1
3
241
"everything's great" is doing a lot of work here. He's crying over the hot wings: "Great! Everything's great!" The subtext: "I can see my ancestors." Tone carries the part we leave out of the sentence.
1
1
352
your AI agent should know when "thanks" means "send a boat." He's waist-deep in his kitchen: "Wonderful. Thanks." The agent: "So glad I could help!" Perfect transcription. Terrible time to celebrate.
1
1
2
225
her haircut has a wingspan. She says: "I love it!" She means: "Why is it a bird." A transcript gets every word right. Her voice has a different review.
1
1
3
271
a perfect transcript can still miss the point. "I never said she stole my money." Seven words. Shift the emphasis and you change who's accused, what's denied, even what was stolen. We're building speech models that hear how you say it.
2
1
11
546
Emphasis tells a listener what you're contrasting. Context fills in the rest. We're building speech models that keep more of that information. Explore Oruk: oruk.ai
102
we made a speech model 13× smaller by making it repeat itself. 52 layers → 2 shared blocks. 2.04B parameters → 153M. The encoder loops 24 times. The decoder loops 28. Smaller weights, roughly the same compute. @OrukLabs
55
69
13
1,097
54,498
That familiar voice might be AI. Meet OrukZero, our AI speech detector. Upload an English voice clip. Get a human-or-AI prediction and segment scores for longer recordings. Try the demo free. No account needed. API included with paid Oruk plans. Links below.
1
3
7
644
Try OrukZero free, no account needed: oruk.ai/orukzero API docs: oruk.ai/docs/synthetic-speec… API access is included at no extra cost with active paid Oruk plans. Free trials and prepaid-only accounts aren't eligible.
224
hearing is no longer believing. Meet OrukZero, our AI speech detector. Upload an English voice clip. Get a human-or-AI prediction. Try the demo free. No account needed. API included with every paid Oruk plan. Built by @OrukLabs. Link below.
4
3
1
25
2,331