The hard problem isn't the codec. It's whether the samples arrive.
“Lossless” is a promise about the codec. It says nothing about whether the audio arrives. Those are two different guarantees — and it's the second one your client hears.
Uncompressed audio is table stakes here, and every product that does it says so. It means no perceptual coding and bit-exact samples, which is a real thing and worth having.
What it doesn't tell you is whether those samples arrive. A stream can be perfectly uncompressed and still gap every time the link hiccups — and a gap is the thing your client remembers. You won't hear it, because you're monitoring your own DAW and it sounds fine from where you're sitting. They will.
Kormix is built around the second promise. Here's how.
A session gets its own gateway on high-performance cloud infrastructure — started when you start it, destroyed when you finish. Not a shared server you queue behind, and not a peer-to-peer connection that works or doesn't depending on what your router thinks of their router. You choose roughly where it runs; we suggest the closest.
Nothing on that path re-encodes your audio. What leaves your DAW is raw PCM — the same samples, up to 32-bit float and 96 kHz. It's encrypted before it leaves your computer, and the gateway passes it through byte for byte: it doesn't decode it, re-encode it, mix it down or store it. No codec gets an opinion about your mix — not in the plugin, not on the server, not in the browser.
Video and talkback travel separately, on purpose. A camera stream having a bad second can't take anything from the mix, because it isn't on the same road.
Every link loses packets. Home broadband, hotel wifi, a café, a transatlantic hop at eight in the evening. This isn't a fault you can shop your way out of — it's what the internet is.
So the question is never does it lose packets. It's what your software does about it.
Give up on it
A packet that hasn't arrived by the time it was due is abandoned, and playback carries on without it. Sometimes the gap is concealed, sometimes you hear it. Either way those samples are gone for good.
Ask for it again
The receiver notices what's missing, requests it, and keeps requesting until it arrives. Nothing is abandoned. The replacement lands inside the reserve it was already holding, and you hear nothing at all.
The receiver holds a small amount of audio in hand before it plays anything — a few tens of milliseconds. That reserve isn't caution: it is the room a re-request needs to complete in. When something goes missing, the round trip to ask for it and get it back is typically a fraction of that reserve, so you hear nothing. When the link is having a genuinely bad moment, the reserve is what gets spent — and if it runs out, playback waits. What never happens is the audio being thrown away.
The same goes for a packet that arrives damaged. Encrypted audio carries an integrity check, so corruption is detectable — and the usual thing to do with it is discard it, leaving a permanent hole from a packet that actually arrived. Kormix treats a damaged packet exactly like a missing one, and asks for it again.
All of that has a price, and it's the same price every time: asking for a packet again takes time, and time is latency. You spend latency for two reasons, and only two — the distance to the person you're working with, and the room a repair needs to complete in.
You choose a preference, not a number. From tight, for a short and stable link, to bulletproof for a bad one. Kormix measures the round trip when you connect, sizes the buffer from it, and re-checks as conditions change. There is no jitter buffer field, no retransmit multiplier and no advanced panel.
Indicative figures
| Setting | Best for | Same country | Intercontinental |
|---|---|---|---|
| Balanced | short, stable links where you want it tight | under 100 ms | ~850 ms |
| Safe Default | anywhere, on a link you haven't tested | ~400 ms | ~1.3 s |
| Maximum | bad links, must-not-glitch | ~1.2 s | ~2 s |
Distance is the part you can't configure your way out of — the buffer is sized from the round trip, so a longer link enlarges it whatever you pick. Two further settings sit below Balanced for very short hops, where a repair has no room to complete; the app marks them as such.
These are ear-to-ear figures — what your collaborator hears against what you played, counting the buffering at both ends. That's a stricter measurement than the one-way figure usually quoted for this kind of thing, and we quote it that way on purpose. Your own interface and DAW buffer size move them. Safe is the default because it holds up on links that aren't short or stable, and most links aren't.
They click a link. No install, no plugin, no account, no card — it opens in the browser they already have, on a laptop or on a phone.
What arrives there is FLAC: compressed for the wire, bit-exact on the other side, the same samples you sent. Most browser receivers hand a guest a lossy stream instead; ours is there too, one click away for a connection that needs it, but it's the fallback rather than the default. A guest can also turn on their own camera and microphone and be in the room properly, rather than listening at it from outside. And guests never pay — the person running the room does, on every plan, including the free one.
Video encoding and decoding happen on the graphics hardware, on both macOS and Windows — not on the cores your plugins want. Talkback is a conversation rather than a stem, so it gets echo cancellation, noise suppression and automatic gain, none of which goes near your program audio.
Kormix records each incoming stream separately, not the mix — and before your faders, so muting a channel to concentrate doesn't gut its recording. Drag the take straight into your DAW. If someone drops off and comes back, their recordings are still there.
Encrypted on both legs. Your session runs on a gateway that exists for it alone and is destroyed when it ends. No audio or video is stored anywhere — there's nothing to leak because there's nothing kept.
Download Kormix, start a session and send someone the link. The free tier is a real tier, not a demo — and the audio path is identical on every plan.