Hands-free, real-time documentation for wet lab scientists. biorigor lives at the bench — capturing every step, reading, and observation as your experiment unfolds, without ever asking you to stop and type.
Documentation that keeps up with the bench, not the other way around.
Hands-free capture
Speak as you work. Audio records in 30–60 second chunks that upload as you go, so a dropped connection costs you nothing and recording never waits on the network.
Every event cites its source
Nothing is paraphrased into a black box. Each extracted step links back to the exact transcript fragments it came from, with the original audio to replay. Uncited events are rejected outright.
Protocol-aware vocabulary
Upload your protocols and papers — PDF, DOCX, XLSX. biorigor pulls out the reagents, instruments, cell lines and gene names inside them and tunes transcription toward the terms your bench actually uses.
An entry you can paste
A reviewed session becomes a plain block of prose with timestamps, reading the way a notebook entry reads rather than like a transcript. Only the events you confirmed go into it. Copy it straight into whatever ELN you already keep — there's a PDF of the full cited review too, for your own records.
Three steps, and only one of them needs your hands.
Record
Start a session and talk through your experiment the way you'd narrate it to a colleague standing next to you. No typing, no stopping, no gloves off.
Review
biorigor transcribes the session, extracts the steps, readings and observations, and lays each one out beside the audio it was drawn from.
Paste
Correct anything that needs correcting, confirm the rest, then copy the finished entry into your notebook of record. What you didn't confirm stays out of it.
Your corrections never overwrite the original. Every edit is stored separately from what the model first produced, so the full provenance of a record — what was heard, what was extracted, and what a human changed — stays intact and auditable.
The things people ask first.
How is this different from a meeting notetaker like Granola?
Granola is good at what it's built for: you're at a laptop, in a call, and it takes your computer's audio and turns it into a summary with action items. A bench doesn't look like that. There's no call to listen to, your hands are in gloves, and the phone is in your pocket a metre from a running fume hood.
The deeper difference is what a mistake costs. When a meeting notetaker mishears a word you get an awkward sentence. When 18.2 g is written down as 80.2 g you have a different experiment — and you may not find out for six months. So biorigor is built on the assumption that it will mishear things: it biases transcription toward the reagents, cell lines and instruments in protocols you've uploaded, every extracted step cites the audio it came from, and nothing reaches your notebook until you've checked it against what was actually said.
And a lab record has to survive being questioned later. Your corrections are stored alongside the model's original output rather than replacing it, so the record holds what was heard, what was extracted, and what a human changed. When someone finally asks why a number looks odd, the answer is still in there.
Do I have to talk in a special way?
No. Narrate the way you would to a colleague standing next to you — "adding 200 microlitres of the PE-LNP prep to well B4, cells look about eighty percent confluent." There's no command syntax, no wake word, and nothing to memorise.
Repeating yourself, thinking out loud, or correcting yourself mid-sentence is fine. Extraction reads whole fragments rather than listening for keywords, so "make that twelve, not twenty" lands the way you'd expect.
What happens when it mishears a reagent or a number?
It will, and the design takes that as a given. A general transcription model has never heard of your construct, and reagent names and numbers are exactly where it struggles.
Two things narrow the gap. Upload your protocols and papers, and biorigor pulls the reagents, instruments, cell lines and gene names out of them and tunes transcription toward the terms your bench actually uses. Then the review step lays each extracted step beside the audio it came from, so confirming a number is a click rather than a hunt through a recording. What you correct is stored separately from what the model first produced, so the record keeps both.
Does it replace my ELN?
No, and it isn't trying to. Your ELN is the system of record. biorigor's job is to get what happened at the bench into it without you typing at the bench.
Once you've reviewed a session, it writes a plain block of prose with timestamps — reading the way an entry reads, not like a transcript — that you copy and paste into whatever notebook you already use. Only the events you've confirmed go into it.
The wifi in my lab is terrible. Does that break it?
It's assumed, not hoped against. Audio records in 30–60 second chunks that upload as the session runs, and capture never waits on the network — that separation is a rule the app is built around, not a fallback. Walk into a cold room or a tissue culture hood and recording carries on; the chunks queue and go up when you're back in range.
Who can see my recordings?
You, and the team building biorigor. Sessions, uploads and vocabulary are scoped to your account, so you see your own work and not anyone else's. We can look at what testers record, because reading back where the app misheard something is how it gets better — and documents you upload for vocabulary can be marked private.
There's a whole page on exactly what gets captured, where the audio goes, and how to say no: what we record.
Can I use it? What does it cost?
It's in development, running with a small number of design partners rather than open to sign-ups, and there's no pricing yet. If your lab is a fit and you'd like to be one of them, email deepak@draftor.bio — a person reads it.
We're heads-down building. Full details are coming soon.