Search...Search plugins and themes...
⌘K
Sign in
  • Get started
  • Download
  • Pricing
  • Enterprise
  • Account
  • Obsidian
  • Overview
  • Sync
  • Publish
  • Canvas
  • Mobile
  • Web Clipper
  • CLI
  • Learn
  • Help
  • Developers
  • Changelog
  • About
  • Roadmap
  • Blog
  • Resources
  • System status
  • License overview
  • Terms of service
  • Privacy policy
  • Security
  • Community
  • Plugins
  • Themes
  • Discord
  • Forum / 中文论坛
  • Merch store
  • Brand guidelines
Follow us
DiscordTwitterBlueskyThreadsMastodonYouTubeGitHub
© 2026 Obsidian

Omni Viewer

battlecookbattlecook9k downloads

View audio, image, video, archive, csv, DBC, ARXML, A2L, ASC, BLF, MF4, Avro, ROS bag, PCAP, PCAPNG, STEP, DB3, ReqIF, excel, word, powerpoint, parquet, Shapefile, HDF5, MAT, jsonl, yaml, toml, markdown, mermaid, PlantUML, hwp, PSD and PDF files inside Obsidian.

Add to Obsidian
  • Overview
  • Scorecard
  • Updates18

Obsidian port of the vscode-omni-viewer extension. View (and in some cases edit) a wide range of file formats directly inside Obsidian.

Supported formats

Category Extensions
Archive zip, rar, 7z, dmg, jar, apk, tar, tgz, gz, tbz2, bz2, txz, xz
Audio mp3, wav, pcm, aiff, aif, aifc, amr, awb, ogg, flac, ac3, aac, m4a
Video mp4, mts, m2ts, avi, mov, wmv, flv, webm, mkv
Image jpg, jpeg, png, gif, bmp, webp, svg
Tabular csv, tsv, xlsx, xls, parquet, arrow/feather (Arrow IPC / Feather V2), jsonl/ndjson/jsonlines
Font ttf, otf, woff
Plain text txt
Medical image dcm, dicom
Email eml
3D mesh glb, gltf, stl, obj
Automotive / measurement dbc, arxml, a2l, asc, blf, mf4, avro, bag (ROS), mcap, db3 (SQLite), reqif, pcap, pcapng, stp/step, h5/hdf5/he5, mat
Web / network html/htm, har (HTTP Archive)
Jupyter Notebook ipynb (nbformat v4)
eBook epub
XML xml, xsd, xsl/xslt, wsdl, xlf/xliff, rss, atom, xaml, kml, gpx, resx, csproj, vcxproj, nuspec
API specification / test report OpenAPI/Swagger JSON or YAML; JUnit XML
Documents pdf (view + annotate/merge/save), docx/doc, ppt/pptx, hwp/hwpx, psd, md/markdown, tex/latex/ltx
Data / source safetensors, gguf, onnx, tflite/lite, keras (and Keras HDF5 models saved as h5/hdf5), mlmodel/mlpackage (Core ML), pte (ExecuTorch), pt2 (PyTorch Export), OpenVINO IR (xml + bin), npy/npz (NumPy), json, yaml/yml, toml, proto, mmd/mermaid, puml/plantuml/iuml
GIS shp (Shapefile), geojson

Features carried over from the vscode extension

  • Audio player with waveform/spectrogram (WASM-accelerated analysis for large files), regions, and region export.
  • Image viewer/editor with filters and annotation tools; filtered images can be saved into the vault.
  • PDF editor: text, stamps, signatures, page reorder, merge with another PDF, save / save-as.
  • CSV editor: cell editing writes back to the file.
  • HAR failure analysis: filter and sort requests by method, status, resource type, domain, URL, headers, and optional body content; inspect request/response details and page metadata alongside summary metrics and a timing waterfall.
  • HTML preview: switch between a sandboxed preview and read-only source. Inline styles and embedded images are preserved; scripts, external resources, navigation, and active embedded content are blocked.
  • Jupyter Notebook preview: Markdown and KaTeX math, highlighted code cells, execution counts, saved text/HTML/image/JSON outputs and error tracebacks, attachments and relative vault images, cell search, and source/output visibility controls. Read-only: code and interactive widgets are never executed.
  • LaTeX structure and math preview: outline, source editing, tables, theorem blocks, KaTeX formulas, and vault-backed \\input/\\include resolution. It is a partial preview, not a TeX compiler or typeset PDF.
  • ML model inspection: ONNX, TFLite/LiteRT, and Core ML graph topology with searchable operator/tensor/IO tables and node inspection; ExecuTorch program instruction graphs with delegate, value, and segment inspection; PyTorch Export (.pt2) graphs with signatures, weights, constants, and module stacks; OpenVINO IR topologies with constant previews from the neighbouring .bin; Keras layer tables, per-layer configuration, and parameter counts; GGUF metadata, tensor index, and quantization summary read without touching the tensor payload.
  • EPUB reader: chapter navigation from the book's table of contents, text scaling, footnotes and in-book links, and a book-information panel. Chapters are sanitized before display; scripts in a book never run.
  • XML inspection: element tree with attributes and namespaces, the source with the selected element highlighted, XPath copy, and search over names, values, and comments.
  • Font inspection: local TTF/OpenType/WOFF specimens, editable sample text, size and variable font axis controls, Unicode character search, paginated glyph previews, and metadata/license details. Fonts are loaded only for the active viewer and released when it closes; WOFF2 and font collections are not supported.
  • Arrow / Feather tables: Arrow IPC files and streams, Feather V2 with bundled LZ4/ZSTD decoding, searchable/sortable/paginated tables, schema and custom metadata, JSON preview, TSV/JSON copy, and JSON export into the vault. Preview is bounded to 10,000 rows; files must fit the 256 MB input limit. Feather V1 is not supported.
  • GeoJSON maps: offline SVG maps for all seven geometry types, polygon holes, feature selection with property inspection, pan/zoom and keyboard navigation. Font and GeoJSON vault reads are limited to 64 MB.
  • Plain text: .txt opens as literal text with whitespace preserved, UTF-8 / BOM-aware UTF-16 decoding, soft wrapping, and an 8 MB preview. Vault reads are limited to 64 MB.
  • DICOM images: native grayscale/RGB image previews, frame navigation, windowing, inversion, zoom and searchable metadata. Compressed images retain their metadata and show a preview limitation. Input limit: 256 MB.
  • JUnit reports: XML reports open with test outcomes, suite/status/search filters, failure traces and output, copy support, and report source. Input limit: 64 MB.
  • OpenAPI / Swagger: JSON/YAML specifications open with operations, parameters, local references, requests/responses/examples, authentication, schema inspection and source. Supports Swagger 2.0 and OpenAPI 3.0–3.2; input limit: 10 MB.
  • Email: EML headers, decoded text and sandboxed offline HTML, inline CID images, raw source, text copy, and attachment saving into the vault. Input limit: 32 MB.
  • MCAP recordings: topics/channels, schemas, message samples with text/hex payloads, metadata, attachments and chunks, with offline LZ4/ZSTD decoding, filters and JSON copy. Input limit: 256 MB.
  • 3D meshes: GLB/glTF 2.0, STL and OBJ with orbit/pan/zoom, fit, wireframe, grid/axes, geometry statistics and glTF animation playback. Requires WebGL. Three.js is bundled; relative buffers/images resolve only inside the model's vault folder. Limits: 128 MB per model, 64 MB per sidecar, 256 MB total sidecars. Extra Draco/KTX2/Meshopt codecs are not bundled.
  • NumPy arrays: .npy arrays and .npz collections with dtype/shape details and a value grid; object arrays are listed but never unpickled.
  • Content-signature rerouting: files whose content doesn't match their extension are opened with the right viewer automatically.
  • Share: review an upload confirmation, then encrypt a copy of the selected file on your device and upload it to Omni Viewer's external share service (max 10 MB, 5-minute link expiry). Choose Upload and share to create and copy the link, or Cancel to stop before uploading. Anyone with the link can view the file. Open a shared link to download it into the omni-viewer-shared vault folder. See the Privacy Policy for details.
  • Refresh command to re-render the active viewer.

How viewers are activated

Obsidian core already handles some extensions (md, pdf, images, audio, video, …). For those, this plugin cannot take over the default view; use the file context menu → Open with … Viewer instead. All other extensions (csv, ipynb, zip, parquet, dbc, hwp, xlsx, …) open with Omni Viewer by default.

Commands (Cmd/Ctrl-P):

  • Omni Viewer: Refresh viewer
  • Omni Viewer: Share current file
  • Omni Viewer: Open shared link

Installation (manual / development)

npm install
npm test
npm run build

Then copy the following into <vault>/.obsidian/plugins/omni-viewer/:

  • main.js
  • manifest.json
  • styles.css

The viewer templates, bundled JavaScript, and WASM assets are embedded into main.js during npm run build, so no extra asset folders are required.

Enable Omni Viewer in Settings → Community plugins. The same bundle runs on desktop, Android, and iOS.

Mobile support

On Obsidian mobile, Omni Viewer uses vault APIs instead of local filesystem paths. Mobile currently supports:

  • ZIP/JAR/APK, audio/image/browser-native video, CSV, PDF, HAR, Safetensors, ONNX, TFLite/LiteRT, Keras, Core ML, ExecuTorch, PyTorch Export, OpenVINO IR, NumPy, JSON/JSONL, YAML, TOML, DBC, XML
  • HTML, Jupyter Notebook, EPUB, Markdown, LaTeX, Mermaid, PlantUML, Protocol Buffer schemas
  • XLS/XLSX, DOCX, HWP/HWPX, PPT/PPTX

Keras models open as .keras archives on mobile. Legacy Keras HDF5 models (.h5) rely on content-signature rerouting, which is a desktop-only path. Keras models are read into memory in full, so the 512 MB limit applies on both platforms; a legacy Keras HDF5 model above it stays with the HDF5 viewer, which reads metadata through the filesystem and opens at any size.

A .mlpackage is a directory bundle, so Obsidian only opens one that has been archived into a single file; a bare .mlmodel opens directly. Core ML models are read into memory in full on both platforms, so the 512 MB limit applies to each encoding.

ExecuTorch programs (.pte) are read into memory in full on both platforms, so the 512 MB limit applies; segment payloads (constants, delegate blobs) are located but never decoded. Separate .ptd data files are not opened.

PyTorch Export packages (.pt2) and NumPy files (.npy, .npz) are read into memory in full on both platforms, so the 512 MB limit applies. On desktop, a renamed .npy, a ZIP holding only .npy members, and a ZIP carrying a torch.export program are rerouted to these viewers by their contents.

XML documents open in the XML viewer, limited to 64 MB on both platforms. .arxml and .reqif keep the automotive viewers. XML saved under another viewer's extension (an Excel 2003 XML .xls, say) keeps that viewer.

HTML documents (.html, .htm) open on desktop and mobile, with a 64 MB vault-read limit. Core displays a bounded source preview above 8 MB or when markup is too dense to preview. UTF BOMs and early HTML charset declarations are supported. On desktop, recognizable HTML under an unknown extension also opens in the HTML viewer; extensions claimed by other viewers keep their existing routing.

OpenVINO IR shares the .xml extension, so an IR is recognized by its <net> root: desktop reroutes it to the OpenVINO viewer, and on mobile the XML viewer hands it over (an .xml up to 512 MB is read there to find out). A same-named .bin in the same folder is read for constant previews and range checks when it is 512 MB or smaller; the topology renders without it.

EPUB books are read into memory in full on both platforms and are limited to 256 MB; chapters and images are inflated on demand. On desktop, an EPUB under another non-archive extension is rerouted by its OCF mimetype member; one renamed to .zip stays in the archive viewer.

HAR files are read into memory in full on both platforms and are limited to 256 MB. For large captures, the viewer retains up to 20,000 requests and bounded request/response body previews; the core viewer reports when content was truncated.

Jupyter Notebooks (.ipynb) use nbformat v4 and are limited to 64 MB on both platforms. The core bounds previews to 2,000 cells, 10,000 saved outputs, 4 MB per preview, and 16 MB total preview content, with warnings when content is truncated. HTML/SVG outputs are sanitized, remote images are blocked, and relative images resolve inside the notebook's vault folder. On desktop, a v4 notebook saved as JSON also opens in the Notebook viewer when its nbformat and cells keys are present in the 64 KB detection window.

GGUF is desktop-first: models are inspected through filesystem range reads, so on mobile only files below 512 MB open (they have to be read into memory in full).

Files created by Save As, PDF merge selection, and archive extraction stay inside the current vault. Desktop-only native helpers are intentionally unavailable on mobile: RAR/7z/DMG/system-tar extraction, ffmpeg transcoding, LibreOffice PDF fallback, and legacy DOC rendering. Browser codec support can differ between Android and iOS.

Architecture notes

Each viewer is an Obsidian FileView. Core-backed viewers, including Jupyter Notebook, mount omni-viewer-core directly into the view and receive vault, navigation, and other host services from a small adapter. Notebook and KaTeX styles are scoped and bundled into styles.css. Legacy viewers host the vscode extension's self-contained HTML templates in an iframe; a bridge emulates acquireVsCodeApi() and handles messages with Obsidian/vault APIs. Both paths map theme variables to the active Obsidian theme.

License

MIT

HealthExcellent
ReviewCaution
About
View and edit dozens of file types directly inside Obsidian—archives, audio, video, images, spreadsheets, PDFs, GIS, automotive logs, and many data formats. Use a waveform/spectrogram audio player, image editor with filters/annotations, PDF editor (annotate/merge/save), CSV editor that writes back, content-signature rerouting, and file sharing.
FilesPDFImages
Details
Current version
0.13.1
Last updated
2 weeks ago
Created
6 months ago
Updates
18 releases
Downloads
9k
Compatible with
Obsidian 1.4.0+
Platforms
Desktop, Mobile
License
MIT
Report bugRequest featureReport plugin
Sponsor
GitHub Sponsors
Author
battlecookbattlecookbattlecook
omni-viewer-web.web.app/
GitHubbattlecook
  1. Community
  2. Plugins
  3. Files
  4. Omni Viewer

Related plugins

Text Extractor

A (companion) plugin to facilitate the extraction of text from images (OCR) and PDFs.

Pixel Perfect Image

Pixel perfect 100% image resizing, copy to clipboard, show image in Finder/Explorer, and much more.

Notebook Navigator

A better file browser and calendar inspired by Apple Notes, Bear, Evernote and Day One.

Claudian

Embeds Claude Code/Codex and other local Agents as AI collaborators in your vault.

Advanced URI

Control everything with URI.

Local REST API with MCP

Unlock your automation needs by interacting with your notes over a secure REST API.

ChatGPT MD

A seamless integration of ChatGPT, OpenRouter.ai and local LLMs via Ollama into your notes.

Quick Explorer

Perform file explorer operations (and see your current file path) from the title bar, using the mouse or keyboard.

Karpathy LLM Wiki

Karpathy's LLM Wiki implementation plugin for Obsidian - turns notes and PDFs into a linked, LLM-powered knowledge base with entity pages, concept pages, graph-powered Q&A, and local-first privacy.

QuickAdd

Quickly add new notes or content to your vault.