Photo-to-Program: help

What the tool does, how to use it, and what it cannot do.

What it does

You give it a table shell (the blank layout of a clinical study report table) as a photo, a pasted text or a document. It reads the shell, shows you what it read, and only after you agree does it write an R program and a SAS program for that table, run the R program, and put the finished table beside your shell so every number that differs is marked. Everything you did is kept in one download with a tamper-evident record.

The steps

Photograph or upload the shell (and, optionally, a spec page or an old program).
Check the reading. Correct anything the tool misread. Nothing runs until you confirm.
Check the derivations it proposes for each row.
Generate and run. The R program runs here against the study's analysis datasets.
Compare the output with the shell, then sign and download the package.

Try it first

Press Try the demo in the header. It runs a bundled photo of a filled table from a made-up study through every step. Its data is invented; nothing in it is from a real trial.

Limits you should know

Demo data mismatches any real study. Until you upload your study's analysis datasets, programs run on invented demo data, so their numbers will not match a photo from a real study.
The reading can be wrong. That is why the check steps exist. Read them; do not click through.
Only the R program is run here. The SAS program is written but not executed on this server.
Do not upload real patient data or confidential sponsor material.

If something goes wrong

Error messages include a short reference. Quote it when you report the problem so the matching server log can be found. The full guide, including the API, is under Help in the studio's left rail.