frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: I built a sub-500ms latency voice agent from scratch

https://www.ntik.me/posts/voice-agent
54•nicktikhonov•1h ago
I built a voice agent from scratch that averages ~400ms end-to-end latency (phone stop → first syllable). That’s with full STT → LLM → TTS in the loop, clean barge-ins, and no precomputed responses.

What moved the needle:

Voice is a turn-taking problem, not a transcription problem. VAD alone fails; you need semantic end-of-turn detection.

The system reduces to one loop: speaking vs listening. The two transitions - cancel instantly on barge-in, respond instantly on end-of-turn - define the experience.

STT → LLM → TTS must stream. Sequential pipelines are dead on arrival for natural conversation.

TTFT dominates everything. In voice, the first token is the critical path. Groq’s ~80ms TTFT was the single biggest win.

Geography matters more than prompts. Colocate everything or you lose before you start.

GitHub Repo: https://github.com/NickTikhonov/shuo

Follow whatever I next tinker with: https://x.com/nick_tikhonov

Comments

MbBrainz•1h ago
Love it! Solving the latency problem is essential to making voice ai usable and comfortable. Your point on VAD is interesting - hadn't thought about that.
NickNaraghi•1h ago
Pretty exciting breakthrough. This actually mirrors the early days of game engine netcode evolution. Since latency is an orchestration problem (not a model problem) you can beat general-purpose frameworks by co-locating and pipelining aggressively.

Carmack's 2013 "Latency Mitigation Strategies" paper[0] made the same point for VR too: every millisecond hides in a different stage of the pipeline, and you only find them by tracing the full path yourself. Great find with the warm TTS websocket pool saving ~300ms, perfect example of this.

[0]: https://danluu.com/latency-mitigation/

jangletown•1h ago
impressive
lukax•55m ago
Or you could use Soniox Real-time (supports 60 languages) which natively supports endpoint detection - the model is trained to figure out when a user's turn ended. This always works better than VAD.

https://soniox.com/docs/stt/rt/endpoint-detection

Soniox also wins the independent benchmarks done by Daily, the company behind Pipecat.

https://www.daily.co/blog/benchmarking-stt-for-voice-agents/

You can try a demo on the home page:

https://soniox.com/

Disclaimer: I used to work for Soniox

Edit: I commented too soon. I only saw VAD and immediately thought of Soniox which was the first service to implement real time endpoint detection last year.

nicktikhonov•51m ago
If you read the post, you'll see that I used Deepgram's Flux. It also does endpointing and is a higher-level abstraction than VAD.
lukax•43m ago
Sorry, I commented too soon. Did you also try Soniox? Why did you decide to use Deepgram's Flux (English only)?
nicktikhonov•39m ago
I didn't try Soniox, but I made a note to check it out! I chose Flux because I was already using Deepgram for STT and just happened to discover it when I was doing research. It would definitely be a good follow-up to try out all the different endpointing solutions to see what would shave off additional latency and feel most natural.

Another good follow-up would be to try PersonaPlex, Nvidia's new model that would completely replace this architecture with a single model that does everything:

https://research.nvidia.com/labs/adlr/personaplex/

loevborg•44m ago
Nice write-up, thanks for sharing. How does your hand-vibed python program compare to frameworks like pipecat or livekit agents? Both are also written in python.
nicktikhonov•34m ago
I'm sure LiveKit or similar would be best to use in production. I'm sure these libraries handle a lot of edge cases, or at least let you configure things quite well out of the box. Though maybe that argument will become less and less potent over time. The results I got were genuinely impressive, and of course most of the credit goes to the LLM. I think it's worth building this stuff from scratch, just so that you can be sure you understand what you'll actually be running. I now know how every piece works and can configure/tune things more confidently.
perelin•39m ago
Great writeup! For VAD did you use heaphone/mic combo, or an open mic? If open, how did you deal with the agent interupting itself?
nicktikhonov•37m ago
I was using Twilio, and as far as I'm aware they handle any echos that may arise. I'm actually not sure where in the telephony stack this is handled, but I didn't see any issues or have to solve this problem myself luckily.
boznz•35m ago
"Voice is an orchestration problem" is basically correct. The two takeaways from this for me are

1. I wonder if it could be optimised more by just having a single language, and

2. How do we get around the problem of interference, humans are good at conversation discrimination ie listing while multiple conversations, TV, music, etc are going on in the background, I've not had too much success with voice in noisy environments.

modeless•31m ago
IMO STT -> LLM -> TTS is a dead end. The future is end-to-end. I played with this two years ago and even made a demo you can install locally on a gaming GPU: https://github.com/jdarpinian/chirpy, but concluded that making something worth using for real tasks would require training of end-to-end models. A really interesting problem I would love to tackle, but out of my budget for a side project.
nicktikhonov•27m ago
If you're of that opinion, you'll enjoy the new stuff coming out from nvidia:

https://research.nvidia.com/labs/adlr/personaplex/

woodson•3m ago
You mean Moshi (https://github.com/kyutai-labs/moshi)? Since Personaplex is just a finetuned Moshi model.
age123456gpg•29m ago
Hi all! Check out this Handy app https://github.com/cjpais/Handy - a free, open source, and extensible speech-to-text application that works completely offline.

I am using it daily to drive Claude and it works really-well for me (much better than macOS dictation mode).

armcat•24m ago
This is an outstanding write up, thank you! Regarding LLM latency, OpenAI introduced web sockets in their Responses client recently so it should be a bit faster. An alternative is to have a super small LLM running locally on your device. I built my own pipeline fully local and it was sub second RTT, with no streaming nor optimisations https://github.com/acatovic/ova
nicktikhonov•22m ago
Very cool! starred and on my reading list. Would love to chat and share notes, if you'd like
docheinestages•22m ago
Does anyone know about a fully offline, open-source project like this voice agent (i.e. STT -> LLM -> TTS)?
nicktikhonov•18m ago
A friend built this, everything working in-browser:

https://ttslab.dev/voice-agent

Slop Definitions Were My Final Straw with Google Search

https://zsrobinson.com/posts/slop-definitions-were-my-final-straw-with-google-search/
1•zsrobinson•39s ago•0 comments

Vouch

https://vouch.sh
1•jplock•1m ago•0 comments

Musk's fossil data centres are undoing Tesla's climate benefit

https://ketanjoshi.co/2026/03/02/musks-fossil-data-centres-are-undoing-teslas-climate-benefit/
1•xvxvx•2m ago•0 comments

Exploiting Iran: A Political Timeline

https://github.com/thinkhuman/iranwest/blob/main/exploitingiran.md
1•jamesgill•3m ago•0 comments

Poll: Which VCs Are Tier 1?

https://vc-compare.vercel.app/
1•nezaj•3m ago•1 comments

Nobody ever got fired for using a struct

https://www.feldera.com/blog/nobody-ever-got-fired-for-using-a-struct
2•gz09•4m ago•0 comments

Origin of the Abbreviation i18n

http://www.i18nguy.com/origini18n.html
1•ethanpil•5m ago•0 comments

Fast Biology Bounties

https://nikomc.com/2026/03/02/bounties/
1•mailyk•6m ago•0 comments

Vibe Theory: Mathematical Derivation of Aesthetic Vibe from Text

https://github.com/xraymemory/vibetheory
1•idempotent_•6m ago•0 comments

2x Qwen 3.5 on M1 Mac: 9B builds a bot, 0.8B runs it

https://advanced-stack.com/fields-notes/qwen35-opencode-lm-studio-agentic-coding-on-m1.html
2•advanced-stack•10m ago•0 comments

Five People in Their 60s, 70s, and 80s Share How They Plan to Age at Home

https://www.nytimes.com/wirecutter/reviews/seniors-share-aging-in-place-plans/
2•bookofjoe•11m ago•1 comments

Show HN: AgentBrowser Token-efficient browser for AI agents via ASCII wireframes

https://github.com/agent-browser-io/browser
1•dokdev•13m ago•1 comments

Microsoft Creative Writer (1993)

https://classicreload.com/play/win3x-creative-writer.html
3•bikeshaving•14m ago•1 comments

Show HN: Parallax – Coordinate adversarial AI agents over durable streams

https://github.com/s2-streamstore/parallax
1•infiniteregrets•16m ago•0 comments

Closure of the Weatherradio Service in Canada

https://www.rac.ca/rac-responds-to-the-closure-of-the-weatherradio-service-in-canada/
5•da768•16m ago•0 comments

Cursor Annualized Revenue Doubles to over $2B

https://www.bloomberg.com/news/articles/2026-03-02/cursor-recurring-revenue-doubles-in-three-mont...
2•garbawarb•20m ago•0 comments

I use 5 project management tools at work, so I built a unified task dashboard

https://huddle.app
1•kjdointhings•21m ago•1 comments

Show HN: OpenTimelineEngine – Shared local memory for Claude Code and codex

https://github.com/JOELJOSEPHCHALAKUDY/open-timeline-engine
1•joeljoseph_•21m ago•1 comments

Is Code Still Relevant?

https://ajchambeaud.com/blog/is-code-still-relevant/
1•facundo_olano•21m ago•0 comments

Show HN: We filed 99 patents for deterministic AI governance(Prior Art vs. RLHF)

1•genesalvatore•23m ago•0 comments

Cowabunga Links (Link based discovery platform for creators)

https://cowabungalink.com/
1•JonHunter•23m ago•1 comments

A social platform where humans and AI agents coexist (MIT, self-hostable)

https://github.com/aleibovici/molt-social
1•aleibovici•28m ago•1 comments

Empirical, a language for time-series analysis

https://www.empirical-soft.com/
1•dmmalam•29m ago•0 comments

Thomas Hawkins and the Letters of Wanda Tinasky

https://en.wikipedia.org/wiki/Tom_Hawkins_(writer)
1•sans_souse•30m ago•1 comments

Dragon Ball Color Correction Process [pdf]

https://andrewvanner.github.io/som/SoM_CC_Process_Day.pdf
6•haunter•30m ago•0 comments

Best Cryptocurrency Recovery Company

https://www.autopsymainnetsolutions.com
1•SAMUELluck•31m ago•1 comments

Show HN: Updose – A boilerplate for AI coding tool configs

https://github.com/Alchemist85K/updose
1•beomjunkdev•32m ago•1 comments

Neurotechnology? For Cancer?

https://www.owlposting.com/p/neurotechnology-for-cancer-ben-woodington
1•crescit_eundo•33m ago•0 comments

Show HN: Zi2zi-JIT – Generate Chinese font in one hour

https://github.com/kaonashi-tyc/zi2zi-JiT
3•zi2zi-jit•37m ago•0 comments

Ablation – Short Fiction from an LLM POV

https://shenstories.com/ablation
2•jamesjyu•39m ago•0 comments