Every version is kept. Flip back anytime.

AI organization and tag search

AI identifies output types and tags from each version’s text. Choose images, video, music, sound, text, code or 3D, then combine directories and attributes. Search, favorites and content range remain independent. Manual corrections are preserved. The organizer lets you choose a source and model, inspect progress, pause and resume.

Version chain

A prompt can carry as many versions as you need. Each one has its own name, its own text and its own references, and each records when it last changed — shown as readable relative time like "just now" or "3 hours ago".

Refining a prompt never means overwriting what worked: add a version and keep going, and draft seven sits alongside draft twelve for comparison. Each version's text box has its own copy button in the top-right corner, so you copy the version you want without expanding and selecting first.

Deleting a version asks twice, and shows you thumbnails of the references that would go with it — nothing disappears without you seeing it first.

Every version also carries its own negative prompt (what should not appear) in a small red box under the text, with its own copy button. Text and negative prompt are always copied separately, and both are translated together when you translate a version.

Automatic snapshots

The version chain is what you keep deliberately; snapshots are what the app keeps for you. Before every change is written, 77Fox stores a copy marked as the state at creation, before an edit, or before a deletion, with relative time and the first line of the text so you can spot the right one at a glance.

"Restore an earlier version" in a version's More menu brings it straight back — title, category, favourite status, version name, text and references all return to how they were. Snapshots roll over automatically, need no housekeeping from you, and travel with sync to your other devices.

Reference material

Keep style images, sample clips and layout references with the version they belong to. Each version supports up to 15 attachments, so you can review the text and its references together.

Four ways to bring files in

• Drag files from your file manager onto a list row, a grid card, the editor or an entry box;
• Paste with Ctrl+V — clipboard files and screenshots both work;
• Pick files with the plus button;
• Drag an image straight off a web page.

Images cover PNG, JPG, WebP, GIF and SVG; video covers MP4, MOV, WebM, MKV and more; audio such as MP3, WAV and FLAC works too, as do PDF and PPT. On desktop a single image can be 100 MB and other media 500 MB; on the web the limit is 64 MB per file. Anything rejected tells you why — no file quietly goes missing.

References can be reordered by dragging, copied or saved individually from the right-click menu, or saved as a whole group at once; on desktop you can also drag them straight out of 77Fox into another application.

One-click translation

Prompts often need an English and a Chinese pass, sometimes Japanese or Korean too. Pick "Add version and translate" in a version's More menu, choose the target language, and the translation joins the chain as a new version named after that language, leaving the original exactly as it was.

Targets cover Simplified Chinese, English, Japanese, Korean, French, German, Spanish, Portuguese and Russian.

You choose the route

• The free default channel — nothing to configure, nothing to pay;
• Cloud translation included in your plan, billed against your subscription;
• Your own API key, billed on your own account;
• An offline model on your machine, where not a word leaves your computer.

If the text is empty or the translation returns nothing, the app refuses and tells you why rather than leaving you an empty version.

Translation works on the web too, through the cloud channel included in a plan or with your own API key once the vault is unlocked. The free default channel and the local offline model are desktop only.

List and masonry views

The library has two faces. List view is dense — title, time, version count and cover on one line, expanding in place for editing, which suits text-heavy libraries. Grid view is true masonry: cards follow each reference's own proportions and lead with the picture, which suits finding a prompt by its image.

The toggle sits in the bottom toolbar and your choice syncs to your other devices. Covers can show a single image or a four-up grid. Hover previews and right-click menus are identical in both views, so switching costs you nothing.

A large library stays responsive: covers load only as they actually scroll into view, instead of pulling dozens of images the moment the page opens.

In the grid you can resize covers from 60% to 200% with the slider in the bottom rail or Ctrl / ⌘ + scroll, and each card can show one large cover or a four-up tile.

Hover preview

Hover a row or a card and a preview appears alongside it: every version's text and its references laid out together, up to ten versions at once. The text in the preview is editable and writes back immediately, so you never have to open the editor for a small fix.

Hover a thumbnail inside that preview and the image or video opens at its original aspect ratio, up to 4000×4000 — judging the detail in a reference no longer means opening the file. Move the pointer off the thumbnail and it closes at once.

On desktop the preview is its own window and can extend past the edge of the main interface; on the web it appears as an in-page overlay that behaves the same way.

Search, sorting and bulk export

Past a certain size, what you need is not browsing but finding. Search matches titles, body text and the content of every version, highlighting the matched words in the results so you can see exactly which line surfaced the entry.

Two sort orders: automatically by most recently changed, which floats your working set to the top, or manual, where you drag things into the order you want.

To move a library or share part of it, enter multi-select with “Select” (Ctrl / ⌘ click, Shift for a range, Ctrl / ⌘ + A for everything). The action bar then exports, copies as Markdown, corrects tags or deletes. An export is a plain zip (the .77foxprompts extension is optional) with one prompt.md per prompt plus a media folder; desktop and web both import it, duplicates are detected first and skipped by default.

Across devices, without overwrites

Desktop and web use the same library. Changes are encrypted before they leave your device and then written into your own Google Drive — prompts never pass through 77Fox servers, and the data key is managed through your Google authorization, so there is no second sync password to remember.

What if two devices edit the same prompt? Categories, prompts and versions are compared record by record, and conflicts are kept as records rather than resolved by overwriting one side. If the data in the cloud turns out to belong to a different account, the app stops and tells you instead of writing someone else's library onto your machine.

References are identified by content, so the same image travels between devices only once. Everything works offline and catches up when you reconnect.

Screen text and translation

When you’re collecting text or prompts from images, press Alt+A to extract the text on your PC or lay a translation over the original. Recognition, translation engines and limits are covered under screenshot translation.

Create prompts from images

For AIGC creators organizing visual references: open the image menu while adding or editing a prompt, choose a model and output language, then start or cancel generation. An empty version is filled; existing text is preserved in its version and a new version is added. English is the initial output language, with a separate saved default. Desktop supports your plan, configured APIs, local CLIs and local services; Web supports your plan or configured APIs.