Supported file formats
Any file can go into a workspace — which ones you can edit as text, which are kept as-is, which get extracted text the AI can read, and which formats you can export to.
You can upload any file — nothing is rejected for being an unknown format. The server decides how to store a file from its content (not its extension), then decides which derived data to generate from the format it actually detects.
Text or binary
A file is stored as text, editable in the built-in editor, when all of the following hold; everything else is stored as binary, kept byte-for-byte as uploaded:
- The whole file is valid UTF-8 (with or without a leading BOM) and contains no NUL bytes
- It is at most 10 MB
- Its line endings are consistent (all LF or all CRLF)
- Its extension is not
.pdf/.rtf— those are never edited as text, even when they happen to be plain ASCII
Text files remember their original BOM and line endings and get them back when downloaded or synced to a local folder. Files in GBK or other encodings, and UTF-16 files, are stored as binary with no loss. SVG and HTML are text too: you can edit them, and they also have an isolated preview (their scripts never run inside VizChat's pages).
Formats that get extracted text
Among binary files, these formats get their text and outline extracted so the AI can read and cite them:
| Format | How | Credits |
|---|---|---|
Word / Excel / PowerPoint (.docx .xlsx .pptx), legacy Word / Excel (.doc .xls) | Local parsing | Free |
| The text it carries is extracted for free on upload; the whole file goes through text recognition (OCR) once, the first time it's read | Text recognition per page, charged once per file (see Billing rules) | |
| Images (PNG, JPEG, and others the current OCR service accepts) | OCR once, the first time its text is read; by default the AI looks at the image itself without recognizing text | One page per image when recognized |
Other formats get no extracted text but still get file info (format, dimensions, and so on); zip archives also get a listing of their contents. Legacy PowerPoint (.ppt) uploads and downloads fine but has no extracted text — resave it as .pptx if the AI needs to read it.
Which viewer opens a file
| Content | Viewer |
|---|---|
| Text | Text editor; SVG and HTML open in preview first, with a switch to edit as text |
| Images | Image viewer |
| Audio, video | Player |
| PDF, Office documents | Document preview (shows the extracted text) |
| Anything else | File info card and download |
Size limits
- Web upload: 50 MB per file
- Desktop local-folder sync: 256 MB per file
- Your account's total storage depends on your plan — see Documents → Storage
Export formats
| Artifact | Supported formats |
|---|---|
| Storyboard | .vizpkg — portable archive holding the board plus every file it references |
| Document | Download of the original file, byte-for-byte, under its original filename |
| Chat sessions | .json — all sessions, from Settings → Data Management |
| Image / PDF / video | Not available yet — see Export overview |
Next step
Keyboard shortcuts
All keyboard shortcuts in VizChat — workspace, canvas, Storyboard, chat, file tree.
Local workspace CLI
The `vizchat` command makes a workspace a real folder you can edit — syncing it both ways with the cloud, exporting and importing .vizpkg files, and answering the two questions a plain file tool cannot: which ref is this file, and is this workspace still consistent.