README file from
Github
An Obsidian plugin that bakes images into your notes as self-contained Base64 embeds — and un-bakes them back into regular files whenever you want.
Why would I want this?
Your notes travel as single files. A note with baked-in images is one .md file that carries
its own pictures. Mail it, paste it into a gist, sync it, move it between vaults — nothing breaks,
no attachment folder needs to come along.
Screenshots stop cluttering your vault. If your workflow is screenshot-heavy — meeting notes,
bug reports, study notes, recipes — every paste normally creates another Pasted image 1748….png
in your attachment folder. With Image Baker the screenshot lands inside the note it belongs to,
and nowhere else.
Images live and die with their note. A baked image exists only in the note that uses it. Delete the note and the image is gone with it — no orphaned attachments to hunt down later, no "is anything still using this file?" anxiety.
It's reversible, so there's nothing to commit to. Every conversion works both ways, down to the original file name and folder. Bake a note, change your mind, extract — you're back exactly where you started. Try it on one note or convert a whole vault, with a dry-run preview either way.
What it does
Getting images in
- Paste or drag & drop an image and it is baked straight into the note — no attachment file
is ever created. Dropped files keep their real name; pasted screenshots get a clean
<note name> <timestamp>.pngname. Anything that isn't a supported image within the size limit is left to Obsidian's normal handling. - Convert existing images — wiki embeds (
![[photo.png|300]]) and markdown images () convert in place: per image, per selection, per note, or across the whole vault. - Optional optimization — re-encode to WebP or JPEG (configurable quality, optional downscaling) while baking, typically shrinking screenshots several-fold. Only used when the result is actually smaller; SVG and GIF are never touched.
Getting images out
- Extract any embed back into a real file — original name, sizing, and folder restored, name collisions resolved, duplicate embeds extracted once.
- Copy image puts the pixels back on your clipboard for use in any other app.
Everyday handling
- Right-click works everywhere: a rendered image in the note (extract / copy / reset size / delete), an image link in source mode, an image file in the file explorer (Embed image into notes that use it), or a note in the file explorer (embed/extract all of its images).
- Click an image, run a command: Embed / Extract / Copy selected image act on the image you clicked or the link under your cursor.
- Collapsed embed data — the editor folds the long Base64 text behind a small
base64 · 142 KBpill; click to expand, move away to fold again. - Image list sidebar — every image of the current note with badges, jump-to-image on click, per-image Bake/Extract buttons, and multi-select with batch conversion.
- Batch dialogs show a dry-run summary first ("Found 214 embeddable images (~38 MB) in 96 notes."), report progress, and can be aborted mid-run.
Safe by design
- Source files are moved to the trash, never hard-deleted — and only when no other note still references them. Deletion can be turned off entirely.
- Notes are modified atomically; image links inside code blocks are ignored.
- Oversized images are skipped (1 MB by default, configurable) instead of silently bloating notes.
- Every operation is logged to the developer console at a configurable level (settings → Log level), so unexpected behavior is easy to diagnose.
Settings
| Setting | Default | Description |
|---|---|---|
| Collapse embedded image data | on | Fold Base64 payloads behind a size pill |
| Embed images on paste | on | Bake pasted images straight into the note |
| Embed images on drop | on | Bake dragged-in images straight into the note |
| Delete source files after embedding | on | Trash the original file once baked in |
| Maximum file size to embed (KB) | 1024 | Skip images larger than this (0 = no limit) |
| Optimize images before embedding | off | Re-encode images while baking them in |
| Optimized format / quality / max width | WebP / 75 / 0 | Target encoding for optimization |
| Extracted link style | Wikilink | ![[image.png]] or  |
| Log level | Warnings | Event logging: Off, Errors, Warnings, Info, Debug |
Good to know
- Base64 makes notes larger — roughly a third over the raw image size. The size limit and optimization settings keep this under control.
- Bases / card views don't render baked images — inherent to data-URI images; extract the image back to a file if you need it there.
- Publishing depends on the renderer — Obsidian renders baked images everywhere, but some external markdown renderers strip them (GitHub notably does). Extract before publishing there.
- Excalidraw and Canvas are out of scope — both store image references in JSON, not markdown.
- Mobile paste is untested so far; everything else is platform-neutral.
Installation
From the community plugin list
Once accepted into the community catalog: Settings → Community plugins → Browse, search for "Image Baker".
With BRAT
Install the BRAT plugin and add
bartekmp/obsidian-image-baker as a beta plugin.
Manual
Download main.js, manifest.json, and styles.css from the
latest release and place them in
<your vault>/.obsidian/plugins/image-baker/, then enable the plugin in
Settings → Community plugins.
Developing
git clone https://github.com/bartekmp/obsidian-image-baker.git
cd obsidian-image-baker
npm ci
npm run dev # watch mode; clone into a test vault's .obsidian/plugins/image-baker/
npm test # unit tests
npm run build # production build → main.js
Contributions are welcome — see CONTRIBUTING.md for the branch, commit, and release conventions.
LLM policy
The TypeScript source code in this repository is mainly generated with the LLM assistance.