Bring your spectra. Find your people.
Spectra Insight is an open commons for spectral data. We read the file your instrument already writes — vendor formats parsed directly, and a careful AI reader for the headers we do not recognise. Refine it in a toolbox that records its own steps, ask the questions a manual cannot answer, and publish a citable record your field can genuinely reuse.
Free, and open to every spectroscopist — students included. Sign in with Google, GitHub or ORCID, or look around as a guest first.
Acetaminophen powder · 785 nm
Horiba iHR320 · x50 · 1800 gr/mm · 3 × 15 s
A real measurement, not an illustration — acetaminophen powder straight off a Horiba iHR320. Cropped to the fingerprint region, decimated to keep this page light, and despiked: three steps, each one written down and replayable. That is the whole idea.
Why this exists
There is far more good data than there are papers to put it in
Nothing is wrong with the measurements already on your drive. They simply have nowhere to live that is not a figure in a paper. Four things change when they do:
Work that outlives the project
A PhD ends and four years of spectra can leave with the student, on a lab desktop nobody logs into any more. Nothing was wrong with those measurements — they just had nowhere to go. Deposited here, they keep their author, their provenance, and their usefulness to whoever comes next.
Someone has already seen your peak
You have a band you cannot place. Someone three labs over placed the same one years ago and never wrote it up, because “I assigned a peak” is not a paper. That knowledge is real; here it finally has an address, and you can simply ask.
Processing you can hand to a colleague
“Baseline corrected and smoothed” leaves the next person guessing which baseline, what window, what order. When every step, parameter and algorithm version travels with the result, your method becomes something a colleague can replay exactly.
The null result still counts
The sample that showed nothing, the batch that failed, the reference that did not match — all real evidence, and all unpublishable anywhere else. Post it and you spare the next person the repeat measurement.
Sharing
Publish when you are ready — never before
Everything you upload is private the moment it lands, and stays that way until you decide otherwise. Publishing is a choice you make per record, and it never destroys provenance.
Private drafts by default. Publish under CC BY 4.0 or CC0 when it suits you, set an embargo date if your paper is still under review, and unpublish again if you change your mind.
Every published record gets a citable accession number, so colleagues can point at your data the way they point at a paper.
Link the manuscript and we verify it against Crossref — the journal card carries its real quartile and citation count, not a claim.
Credit where it belongs: co-authors by handle, and ORCID sign-in earns a verified badge on your profile.
Feed and discovery
A feed of real measurements
The home feed mixes published spectra with findings — threads attached to actual records, so an answer points at data instead of at somebody's memory.
Switch between Discover and Following. Follow the people whose work you already rely on and their spectra and findings come to you.
One search box for the whole platform: ⌘K opens a palette across compounds, spectra, findings and people.
Narrow the feed by tag, by author, or to DOI-verified work only when you want the peer-reviewed slice.
Profiles gather someone's posts, their library contributions and a contribution heatmap, so it is easy to see who works on what.
Your data lab
A workspace that keeps its own lab notebook
The Lab is yours: your raw files, your private projects, and a toolbox that writes down what it did while you work — so reproducibility costs you nothing extra.
Upload and we read it. Ocean Insight, Horiba and JCAMP-DX files are parsed in full; Renishaw, WITec and Bruker files are recognised and read for what metadata they expose; and a careful AI reader takes on headers we do not know. You confirm every field before anything is recorded.
Thirteen preprocessing steps, arranged by drag and drop with a live preview: despiking, Savitzky–Golay smoothing, ALS, ModPoly and airPLS baselines, SNV, MSC, vector, area, min–max and peak normalisation, cropping and resampling.
Every pipeline becomes a hash-addressed ledger you can replay, save as a reusable routine, or fork from someone else's spectrum into your own workspace. Your raw file is never overwritten.
PCA and k-means run right in the browser for a first look, and each run records exactly which spectra and which versions went into it.
The shared library
A reference library anyone can read, and anyone can add to
A growing corpus of openly licensed reference spectra, free to browse without an account, drawn from sources whose licences we checked one at a time.
Starting with the Raman Open Database, with more sources added as we clear each one's licence one at a time — a plastics set and a collection of pure solvents and reagents are next in line. Where a licence could not be established we leave the data out entirely, including collections we would have liked to include. See what's actually loaded on the library page itself, not just here.
Search by name, chemical formula, CAS number, SMILES, InChIKey, or a single element symbol — “Fe” returns every iron-bearing formula.
Each entry shows its peak list, its elemental breakdown, its licence and the source it came from. A citation-backed band table covers the assignments themselves.
Identify an unknown against the corpus, or resolve a mixture into per-reference contributions — and add your own published spectra, badged as community entries and matchable straight away.
How it works
From the file on your desktop to a citable record
- 1
Upload
Drop in the file your instrument actually wrote. Vendor parsers handle the formats they know; a guarded AI reader takes on the headers they do not.
- 2
Confirm
You review the parsed metadata before anything is recorded. The machine may suggest. It never decides for you.
- 3
Process
Work through the toolbox in the open. The raw file stays untouched and recoverable; every step is written down as you go.
- 4
Publish
Release a citable record with its provenance intact and its manuscript link genuinely verified — or keep it private indefinitely.
Our commitments
The promises the commons runs on
An open commons only works if the rules are stated up front and do not move. These are ours.
Your raw source file is immutable and always recoverable. We never overwrite what your instrument produced.
You confirm the metadata. AI may suggest — it never silently decides.
Every result we display traces back to its source data and the exact steps that produced it.
Private by default. Publishing is explicit, reversible, and never destroys provenance.
Scientific search answers scientific questions. Social popularity never changes scientific ranking.
The commons gets better every time someone adds to it
Bring one file, one question, or one finding. That is enough to start — and it is enough to help.
Spectra Insight is in early access. Things are still being built, and we would rather tell you that than pretend otherwise. By continuing you agree to our Terms and Privacy Policy.