What it does
You give it a structure — as SMILES, a drawing, a MOL/SDF file or a picture. It checks the molecule, asks the AI model you picked for the spectra you ticked, and shows charts and tables where every peak is tied to atoms.
About
Spectrum prediction that is clear about what it is.
Spectra turns a molecule into predicted ¹H NMR, ¹³C NMR and IR spectra using an AI model of your choice — and always shows them as predictions, never as measurements.
Who it is for
Students, teachers and chemists who want a quick first look at what a structure's spectra might resemble, or a way to practise assigning peaks to atoms.
What it is not
It is not a measurement and not a replacement for one. Predicted values can be wrong, so do not use them as evidence in a paper, a regulatory file or a safety decision.
Under the hood
How a prediction is made
Five stages — the AI is only one of them.
-
1
The structure is checked
RDKit parses the molecule, rejects invalid ones and produces the canonical SMILES, InChIKey, formula and mass. Each atom gets a number you will see again in the results.
-
2
You set the conditions
NMR solvent, spectrometer frequency and IR method are part of the request, because they change what a spectrum looks like.
-
3
The model predicts each spectrum
OpenAI or Claude — your choice — receives the verified atom list and returns a structured answer. Every spectrum is a separate request, so one failure never blocks the others.
-
4
The answer is checked against the molecule
Peaks that point at atoms which do not exist are dropped, equivalent carbons are merged, and anything the model left out stays empty instead of being filled with a guess.
-
5
You explore the result
Charts, peak tables and atom highlighting, all marked as predicted. Click a peak to see the atoms behind it.
Principles
What we hold ourselves to
Predicted, never measured
Every result says it is a prediction.
No invented numbers
Missing values show as “Not predicted”.
Chemistry first
The molecule is validated before the AI sees it.
Open to every language
English, Persian and Arabic, left-to-right or right-to-left.
Technology
Built on proven tools
- RDKitStructure parsing, canonical SMILES, InChIKey and the drawings.
- JSMEThe molecule editor you draw in.
- OpenAI · ClaudeThe models that predict the spectra and read structure images.
- Node.js · ExpressThe server that connects them. API keys stay on it.
Your data
Where things go
- In your browser
- Your analysis history, theme and language. It never leaves your device.
- To the AI provider you pick
- The molecule (SMILES), the chosen conditions and, for image input, the picture you upload.
- On our server
- Your name, e-mail and a hashed password if you create an account, and any message you send through the contact form.
Questions, ideas or a prediction that looks wrong?
Tell us — or just try a molecule.