View only. Sign in to run this demo.
Source Normalizer
Transform data (PDFs, Word files, spreadsheets, text, recordings and images) in a uniform Markdown format.
How to use the Source Normalizer
- 1Add the sourcesPDFs, Word files, spreadsheets, text, recordings and images, or start from a ready-made set.
- 2Label themMark each file primary (first-hand evidence) or secondary (background).
- 3NormalizeEach file becomes Markdown that keeps its file name and label. Progress shows file by file.
- 4Explore and use the textsSearch across all texts, read them, fix errors, then download them or send them to the Knowledge Curator.
- Documents are converted locally. Only recordings and images call a model. A scanned PDF with no text layer stops the run, and recordings carry no timestamps for each utterance.
- The result lives in this browser tab. Reloading or closing the tab loses it, so download the .zip first.
Start from a ready-made set of sources
Click primary or secondary to switch a label. Primary sources are your own first-hand evidence, such as site notes or recordings. Secondary sources are background, such as older reports. Not supported: PowerPoint (.pptx), old Office formats (.doc, .xls), scanned PDFs without a text layer.
The language of the speech in the recordings, or a language code of your own.
Leave it empty to use the server's transcription model.
Leave it empty to use the server's model for describing images.
No result yet
Load a ready-made set or add files, then click Normalize. The texts appear here, where you can search all of them at once.
