Markdown ZIP
A Markdown file for each document, with figures in an images folder. Use it in your own document pipeline or AI tools. Best for reusing extracted content.
Annota Studio · Documents & OCR
Turn documents into data your AI can use.
Extract text, tables, and figures from PDFs and scans, bring in your Word and Excel files as they are, and transcribe audio recordings. Review the result, then use it in your own AI tools or ask questions about it in Annota Brain.
From PDF to editable text, tables, and figures.
How it works
Extract. Review. Reuse.
Keep each original next to its extracted content while you prepare your documents.
Upload your documents
Create a document dataset and add PDFs, scans, Word and Excel files, or audio recordings. OCR reads scans page by page, Office files are ready in seconds, and audio becomes timestamped text.
Review the result
Compare the original with the extracted pages. Correct the text, check figures and tables, and approve the document.
Preserve your work
Save a dataset version to keep a stable reference. Export that version or continue working with the current dataset.
Exports
Pick the export that fits.
Both exports include your OCR corrections and extracted figures.
LLM Wiki
A folder you can open in Obsidian, with the original documents, corrected transcriptions, an index, and instructions for an AI agent. Best for giving an AI agent organized, reviewed sources.
Models
Coming soonFrom reviewed documents to your model.
Prepare the reviewed text and domain-specific examples your own language model will build on. LLM fine-tuning will start from this material.
Work from your AI agent
Keep your agent connected to the source.
Through MCP, your AI agent can read your documents, correct the extracted text, approve what is ready, and fetch your exports.
AI agent - Annota AI MCP / Example
Annota Brain
Ask questions across your documents.
The documents you review here power knowledge bases in Brain, and files added in Brain show up here for review. Select a knowledge base in Brain and answers drawn from your documents cite them.
Frequently asked questions
Which files can I upload?
PDFs and scans (PNG and JPEG), Office and text files (Word, Excel, PowerPoint, OpenDocument, RTF, HTML, EPUB, CSV, Markdown, and plain text), and MP3, M4A, and WAV recordings, up to 100 MB each. PDFs can have up to 1,000 pages and text files up to about 250 pages; older DOC, PPT, and XLS files are not supported. Reading scans uses credits for each page and transcription for each started minute of audio, while Office and text files are free to add. The app shows the estimated cost of each file before upload. Video datasets are coming soon.
How does audio transcription work?
Each recording is transcribed in parts of up to twelve minutes into timestamped text that you review and correct like OCR. You also get a free overview of the whole recording. Transcripts flow into dataset versions, exports, and Brain knowledge bases; the overview is not searched or cited.
Does an LLM Wiki export answer questions for me?
The export gives you source files and instructions. You open it with your own compatible AI agent, which then works with those sources. To ask questions inside Annota AI, use Brain: it answers from knowledge bases built on your document datasets and verifies each answer against the source pages.
How do datasets and knowledge bases work together?
Every knowledge base follows exactly one document dataset, marked In Brain in Studio. Turn any document dataset into one with Use in Brain: it indexes only approved documents by default, or all documents with extracted text if you choose. Knowledge bases created in Brain index all their documents. A dataset can't be deleted while it backs a knowledge base.
Do versions and document exports use credits?
No. Saving a version and exporting Markdown ZIP or LLM Wiki use no credits, even at zero balance. Their retained archives count toward storage, so you need enough available space.
How many dataset versions can I keep?
Up to 10 versions per dataset on Free, Pro, and Team. Each version keeps your data exactly as it was. Delete an unused one to make room; versions used to train a model are kept while that model exists.
Where can I check costs and data use?
Pricing explains plans, credits, storage, and usage examples. You can buy credit top-ups on Free, Pro, and Team; purchased credits never expire. Original files, OCR text and figures, transcripts, saved dataset versions, and stored exports all count toward your organization's storage allowance. Privacy & Security explains hosting, providers, access, and how your data is used.
Start with a document.
Create a free account, upload a PDF or scan, and review your first OCR result.