Files get a description you can edit
Every photo, PDF, document and voice note gets a short AI-written description and its text, so you can search for it later — and you can rewrite the description any time.
Every file you add to Kaizendex — a photo, a screenshot, a PDF, a document, a voice note, a video — gets a short description written for it, once, right after you upload it. It's the "what is this" line: a photo of a bowl of ramen on a wooden table, a two-page lease for a flat, a voice note about buying oat milk. You see it under the file, you can rewrite it, and it's what lets you find the file again by what's in it.
What gets written
| You attach | The description says | Also kept for search |
|---|---|---|
| A photo or screenshot | What's in it — the subject, the setting, any visible text, the mood | The text in the picture (a menu, a whiteboard, a receipt) |
| A PDF or a Word / PowerPoint / Excel document | A one-paragraph summary | The document's full text — page by page, slide by slide, sheet by sheet — so a sentence deep inside a 40-page PDF is findable and points at its page |
| A plain text, Markdown, CSV or JSON file | A one-paragraph summary | The file's text |
| A voice note or audio recording | One line: what was said | The full transcript, in 30-second stretches you can jump to |
| A video | What the cover frame shows, plus the length | — (the audio isn't transcribed yet; the file is marked partial) |
| Anything else | The file type and size | — |
How much of each file gets read, and how the AI reads it back, is spelled out in Uploading files and what the AI can read.
The description is at most 60 words and is written in the language of the text in the file (or English when there is none). It usually lands within a minute of the upload; until then the file shows a quiet Describing… line. Nothing waits on it — your entry saves and syncs exactly as before.
Where you see it
- In the viewer. Open any file from a Files & media property (click it on the web, or tap it on your phone). The description sits under the file with a small AI tag when the model wrote it, or you once you've edited it.
- In Files & media. Every row shows its description under the file's name, so you can skim a whole library of photos by what they show. See Files & media.
Editing it
Click or tap the description under the file, type, and press Enter (or tap away) to save. Esc cancels. It saves instantly and syncs to your other devices like any edit.
Once you've edited a description it's yours: the AI never rewrites it. Clearing it entirely is your call too — the file simply shows no description afterwards, and the AI won't fill it back in.
Searching by what's in a file
The description and the extracted text feed the same search your AI uses. So in any chat you can ask:
- "Find the photo of the whiteboard from the planning meeting."
- "Which PDF mentions the deposit amount?"
- "What did I say in the voice note about grandma?"
The AI finds the file through its description and text, then can open the actual picture or page to answer. Files show up in search results as chips that open the entry they belong to. See Chats and AI agents.