Files, Voice Notes & Local OCR

Orbit Thread extracts searchable text from supported files, voice recordings and images locally on your device. Optional sync may transfer permitted content, and BYOK sends relevant excerpts to your chosen provider.

Files and Documents

When you drag or import a file (DOCX, PDF, CSV, JSON, HTML, Markdown, TXT, RTF), Orbit Thread preserves the immutable original file on disk and extracts a searchable text representation locally in the background.

Supported formats

• Documents: PDF (text and scanned), DOCX, ODT, RTF, TXT, Markdown, HTML.
• Presentations: PPTX, ODP — text and speaker notes, searchable by slide.
• Spreadsheets & data: XLSX, ODS, CSV, JSON — searchable by sheet.
• Images: PNG, JPEG, WebP, GIF (BMP and TIFF where your browser can open them) — printed text read with local OCR.
• Audio: MP3, M4A, WAV, WebM, OGG, AAC, FLAC — transcribed locally.
• A single file can be up to 200 MB. Other file types can be attached and kept, but aren’t searchable.

• DOCX & Office: Extracts headings, lists, tables with columns, and links while preserving Unicode and structure.
• PDF: Extracts text page-by-page, retaining page number provenance so search and citations take you directly to the relevant page.
• Tabular Data (CSV & JSON): Preserves headers, column-value associations, and key paths so deep facts can be queried naturally.

Voice Notes & Local Transcription

Click the microphone button to record an idea. A live timer and visual indicator show the recording status. Press Stop to save or Discard to cancel. The original audio recording is preserved locally.

Transcription can run locally in your browser using Whisper via WebAssembly/WebGPU once the required model resources are available. Local transcription does not send audio to an AI provider; if you enable sync, encrypted audio may use Thread’s temporary device relay.

• Timed Segments: Click a timestamp in the transcript to seek near that part of the recording.
• Editable Knowledge: You can edit the transcript at any time. Your edits are saved as a clean user derivative, while the original machine transcript remains permanently intact.
• Offline: Saved recordings and completed transcripts remain available and searchable offline. Starting a new transcription may need a connection if your browser has not retained every model resource; if loading fails, the original recording stays saved and you can retry.

Local OCR for Images and Scanned PDFs

For images (PNG, JPEG, WebP) and scanned or image-only PDFs, Orbit Thread includes a local OCR engine powered by WebAssembly Tesseract with no per-use OCR API fee. It detects text on receipts, documents and slides in your browser.

Editing and deleting notes & files

Every note and file attachment can be edited, renamed, or deleted at any time from its detail page. Deleting moves items to Trash with an option for permanent deletion. Deleting an attachment purges the binary locally and globally propagates a deletion tombstone so no device resurrects it.

Attachment synchronization & temporary relay

Permanent user files live on your devices. When synchronization is active, attachments travel to your other devices through an ephemeral, end-to-end encrypted relay. Each file is encrypted on your device (AES-256-GCM) with its own key, which is locked separately for every receiving device so only those devices can open it. Relay copies are deleted after receiving devices verify storage, or after 7 days.

Status badges show local availability: Local, Syncing, Synced, Waiting for another device, Unavailable on this device, or Sync failed. If a device was offline past the 7-day transfer window, tap Request from another device to pull a fresh encrypted transfer from one of your online devices that has the file.