Workspace

Workspace is the shelf every other module reaches into. Your images and videos live in one Media Library the composer and the studio can open without you leaving what you were doing, and your Brand Voices decide how AI-written captions and carousels sound. The overview lives on Workspace; this is the how.

Before you start#

The Media Library is open on every plan, and nothing in Workspace spends credits: uploads, folders, moves, deletes and brand-voice analysis are all free. How much storage you get depends on your plan, and the meter under the folder list shows what you have used against what you have. Allowances are on the pricing page.

Brand Voices are a paid-plan feature, and your plan decides how many voice profiles you can keep at once. If the create button is greyed out, that is why, and the note beside it links to the plan picker.

No connected social account is needed for any of this: a brand voice reads public websites and public Instagram profiles as they are.

Upload and organise your files#

1. Open Workspace, then Media Library. You land on a grid of your files with the folder rail across the top.
2. Drag files anywhere onto the page, or use the add button in the toolbar. A drop overlay appears while you drag, and a toast reports how many landed. Accepted formats are JPG, PNG, GIF, WebP, MP4, WebM and MOV; the line under the upload button states the per-file size limit, and anything over it is rejected before it uploads.
3. Create a folder from the rail, with a name, a colour and an icon, and it becomes a chip you click to filter the grid.
4. Move files by dragging them onto a folder chip, or by selecting them and using the move action.
5. Select with a click, extend with shift-click for a run of files, or use select all. While anything is selected, a bar offers move, download and delete for the whole set.
6. Narrow the grid with the search box (it matches the displayed name and the original file name), the type filter for photos, videos or GIFs, and the sort menu.

Deleting a folder never deletes what is inside it. The files move to Uncategorized, and the confirmation says so before you commit.

Reuse a file from anywhere#

Two folders exist that you did not create. Anything you save out of the Content Browser, and anything exported from the Content Creator, lands in the folder named after it. Neither can be renamed or deleted, but you can move files out of them.

Everywhere the app needs a file from you it opens the library rather than a file dialog. In the composer and in the studio, a "Choose from Library" button gives you the same grid with a search box; pick one file or several, depending on what the screen accepts, and confirming hands them to the post or the slide. Files dropped straight into the picker upload into the Content Creator folder and come back already selected.

One rule protects the other direction: a file a scheduled or publishing post depends on cannot be deleted out from under it. See Publishing and scheduling.

Teach the AI your brand voice#

1. Open Workspace, then Brand Voices, and start a new voice.
2. Give it a name and a business description. The description is required and does most of the work. Write what you sell, who buys it and how you talk to them, in plain sentences.
3. Add up to five sources: websites or social profiles. Paste a link and the form tells you what it recognised. Instagram profiles are read in depth, profile plus recent posts; other platforms and ordinary web pages are read best-effort.
4. Pick the language the generated text should be written in. If yours is not listed, type it in.
5. Optionally open custom instructions and describe how the brand writes: sentence length, formatting, things to always include or never say.
6. Start the analysis. The work runs in the background, so you can close the dialog and keep working. The voice shows an "Analyzing" badge in the list until it settles, then shows its tone.
7. When it finishes, open the voice and correct what the analysis concluded.

The editor holds the whole profile: tone of voice, point of view, target audience, words to use, words to avoid, emoji usage, language and custom instructions. The analysis fills in tone, point of view and audiences, and seeds the custom instructions with the style it found only when you left that field empty. Your own wording is never overwritten by a re-run. The two word lists and the emoji setting are yours to type; nothing infers them.

Each source row shows its own status and carries its own read button. Re-reading one source refreshes what we captured from that link but does not re-derive the voice; to rebuild tone, point of view and audiences, run the analysis again from the list.

Finally, set a default. The first voice you create becomes one automatically, and generation screens pre-select it. Deleting the default leaves you without one until you pick another.

Reading the results#

The storage meter under the folder rail shows used against total. It is enforced at upload rather than discovered afterwards: an upload that would cross the line fails and stops the rest of that batch. Deleting frees the space as soon as it completes.

Deduplication works on file contents, not file names. Upload a file you already own and the library keeps the copy it has instead of charging your storage twice. Two side effects follow: renaming a file on your computer does not create a second copy, and re-uploading an existing file into a different folder moves the original there.

A brand voice profile is a set of instructions handed to the model on every generation:

• Tone of voice and point of view set the register and who is speaking.
• Target audience tells the model who the copy addresses.
• Words to use and words to avoid go through as lists. Keep them short and specific, because they are a nudge, not a filter.
• Emoji usage decides whether generated text reaches for emoji at all.
• Custom instructions cover shape: sentence length, formatting, structure.
• Language is what the text comes back in, whatever language you wrote the description in.

Only a voice whose analysis completed is applied. One still pending, or one that failed, is ignored rather than half-applied, and the output falls back to your account defaults.

In the composer the order is: a setting you changed on that post wins, then the brand voice, then your account defaults. Until you override a field yourself, switching the voice keeps re-deriving that field from the new one.

What this does not do#

• It does not give free accounts a brand voice. Voices are a paid-plan feature, and storage size follows your plan on every tier.
• It does not read every source equally. Instagram profiles and public web pages work best; a login-walled page or a private profile comes back empty. A voice takes five sources, not a crawl of your whole site.
• It does not generate a thumbnail for an uploaded video. Videos upload, store and publish normally, but the tile shows a placeholder rather than a frame from the clip.
• It does not nest folders. They are one level deep, so lean on names and colours instead of a tree.
• It is not an image editor. Cropping, watermarking and per-platform adjustments happen where the post is composed.

When something looks wrong#

The upload stopped partway with a storage message. You hit the quota, and the batch stops at the file that would not fit; files before it are already in. Delete something you no longer need, or move to a bigger plan, then re-drop the rest.

Deleting a file is refused because a post is using it. The message names how many scheduled posts hold it, so a queued post cannot lose its media. Remove the file from those posts or let them publish, then delete. A bulk delete does not fail on this: it deletes what it can and tells you how many it skipped.

A brand voice is stuck showing failed. Open it and look at the source rows: a link that could not be read is marked as such. Fix the URL or remove it, then run the analysis again. A voice with no readable source still analyses from the business description, so a failure usually means that description was thin and every link unreachable.

You are told to wait before analysing or uploading again. Both are rate-limited per hour, and the message says when to retry. It is a pause, not a failure, and nothing you already did was lost.

Generated copy does not sound like your voice. Check three things in order: the voice finished analysing, it is the voice selected on the screen you generated from, and no per-post override sits on top of it.

Anything else, or the same problem twice, is worth sending to support with the name of the file or voice involved.