Captions and caption direction

Adding, removing, or reordering a mention in the detail editor preserves unrelated unsaved captions and titles, including text typed while the mention saves. Mention footers still follow the selected platform identities. A failed mention save keeps the draft available to retry. Other unsaved caption edits still require Save Changes. Watch-folder rescans preserve existing saved captions, including automatically generated copy; only a new imported post receives its initial automatic caption.

Open a post to edit its individual platform captions and supported titles. Field labels use the same muted sentence-case styling in the inline and detail editors, including Reddit and TikTok titles. Caption Direction adds context for that post when regenerating captions; it complements the saved AI caption style in Settings → Captions. Apply Direction regenerates every platform caption draft, including platforms that share the same caption. A platform’s own Regenerate button updates only that platform. Owner-only Reddit regeneration refreshes body captions while preserving chosen communities and titles. Review generated drafts before saving. A collapsed platform caption may show a hover preview; an expanded caption is edited directly. On an already-posted card, selecting the caption copies it and briefly confirms the copy; the posted detail page also provides a copy control for each platform caption. Generated copy fits the smallest selected caption limit, including mentions and credits. Overlong transcripts ask for a shorter preset. X counts weighted characters, YouTube counts UTF-8 bytes, and TikTok videos use 2,200 characters. Approval blocks overlong edits. Help explains these tools but does not generate captions, code, or unrelated content as a substitute for the product's metered features. Do not claim an edit is saved or applied until the UI confirms it.

The main grid previews the caption for the first selected, connected publishing platform, marked by its platform icon. When no publishing account is assigned and usable, it shows a shared caption without a platform icon, name, caption arrows or platform character limit. Add account remains visible above it. The shared caption can be edited and saved before assigning an account; this does not assign a platform or overwrite saved platform-specific drafts. Posted cards retain their historical platform copy even if an account is later disconnected. On a device with hover, hovering over the preview shows the full caption. Select the preview to edit it in the card, then use the arrows beside the platform name to review or edit each selected platform's caption. Reddit has a post title field, and TikTok photo posts have an optional photo title field. Caption style, Regenerate, and Apply to All are available below the active caption. The style dropdown comes first and starts with the saved photo or video setting. Choose another preset, then Regenerate to preview it for this post. Mentions only is the last option and keeps just that platform’s mentions and credits. The choice is saved with the post, without changing profile defaults. Review the resulting text for each destination before saving; platform-specific handles and names can differ. The count below the caption field and above the action row shows the current caption length and, where applicable, its platform limit. Save Changes commits the edits. While Regenerate is running, Cancel stops that action and keeps the unsaved draft open; once idle, Cancel closes the editor and restores the saved text. Opening an inline caption puts the caret at the end of the existing text. Later taps can place it anywhere. When the mobile keyboard opens, the page adjusts after the keyboard settles, keeping the caption and its Cancel and Save Changes buttons visible when space allows. If the full editor is taller than the available screen, the focused text field takes priority. The image and editor stay within the card, and manual scrolling stops automatic alignment. If the keyboard obscures the actions, scroll the page rather than the image or editor area.

If a caption or title has changed in the main grid, switching views, changing filters, opening another page, or using browser Back shows Unsaved changes. Save & Leave saves the draft before continuing, Discard Changes leaves without saving it, and Stay on Page keeps the draft open. Opening an editor without changing anything does not show the warning. Browser refresh or closing the tab uses the browser's own unsaved-changes prompt.

When no AI caption style is saved, Clear & natural, the first starter, is selected by default. Choose another starter or write a custom style. Typing selects Custom, which remembers your wording while you try other presets. Select Custom to return to it; selecting Clear & natural restores the default. Existing custom styles and migrated standing directions are preserved. The same default applies to new profiles and older blank settings without a bulk settings rewrite.

Generated social copy should be a natural thought or hook, not alt text or a list of visible people, clothing or objects. The generator separates visual-analysis metadata from publishable captions and checks common observer-style description leaks. If a response contains only a detected image description, generation fails without replacing the saved caption or making an automatic extra model request. Review the result before saving. These guards affect newly generated copy, not captions a person has already typed or saved. First-person voice still requires the pictured creator to be recognized in their publishing profile; a self-credit alone does not establish that.

Settings → Captions has separate Photo caption style and Video caption style sections, both with preset chips followed by guidance. Video defaults to Natural summary, with Cleaned-up speech, Transcript, Key takeaways, Short hook, Custom, and Mentions only choices. Transcript uses the available intelligible words without creative rewriting; long transcripts may need shortening for a platform. With no clear speech, including music-only or silent videos, generation and the post dropdown use the saved photo style instead. Until speech analysis finishes, a video shows video options. YouTube has an automatic-description toggle alongside the other platforms.

Settings → Captions has an All Profiles baseline for caption style and Mentions & credit defaults. Individual profiles can override caption and mention defaults independently. Use All Profiles caption settings or Use All Profiles mention settings resets the corresponding section. Shared defaults apply to new profiles and profiles without an override; existing custom choices are preserved. Automatic mention tagging, self-tagging, and Instagram collaborator invitations start off for new users. Including a name or handle in a caption is a separate choice from tagging an account. These preferences affect new posts and newly added mentions, not existing post choices. Save stays visible while scrolling.

On an editable post's detail page, caption and Tags & credits drafts share one bottom action bar with Save Changes and Discard Changes. Saving keeps the detail page open; discarding asks for confirmation before restoring the saved post. The bar stays visible while scrolling and only appears when there are unsaved form changes. The floating Help button stays hidden while these actions are visible. Image tools retain their own draft controls. Approve, Reject, Clear All, and Move to Trash remain visible but are unavailable until unsaved form and image-tool changes are saved or discarded.

Automatic AI captions are editable before publication. One caption uses a shared starting pattern across platforms; Individual lets you choose a separate pattern for each social platform, not for each photo. Mentions appears first among available smart chips. Removing a chip keeps the caret in that same field. Reddit caption and title defaults are visible only to the ShootFeed owner.

Automatically added caption mentions use spaces between handles across platforms. Display-name fallbacks use commas so separate people or credits remain readable. This applies when generating captions or editing mentions, including Reel credit lists. Commas in the creative caption remain intact; existing saved captions change only through the usual edit or regeneration flow.

In the detail editor, turning off the last selected account for a platform immediately hides that platform's caption section. Turning it back on restores the caption. If another account on the same platform remains selected, its shared caption section stays visible. Account selection is separate from turning caption text off.

TikTok and YouTube upload/audience defaults now live in Settings → Publishing, separate from writing preferences. The section and each platform's accordion appear only for an eligible connected account in an active profile. The Default TikTok photo title (optional) remains in Captions because it controls written text, for photos only. Existing saved title patterns remain intact; an empty title stays empty when saved and reloaded. For a connected YouTube channel, Default YouTube Shorts title (optional) provides a separate Smart Text pattern in both generation styles. Use All Profiles or a profile override. Leave it blank to keep suggested titles; custom post titles are preserved. YouTube titles use credited names and stay within 100 characters.

The settings save button says Saved and is muted when there are no unsaved changes. After an edit, it becomes active and says Save Changes; during the request, it says Saving…. Returning every field to its saved value restores Saved without requiring another save.

When a confirmed or confident face match identifies the creator in a photo on their own publishing profile, AI captions speak in first person. Other people remain third person. A photographer credit, title, or uploaded reference photo alone does not establish that the creator appears in the post. Save your name and account handles in Settings → Account → Creator identity, and add reference photos in Face Library. A Photographer can still appear in their own photo; the title describes credits rather than caption perspective. New uploads finish face analysis before generating captions. Existing manually edited captions are preserved; use Regenerate to review a new suggestion.

TikTok is hidden from beta users and the public, including account choices, creator identity, Contacts, Face Library, captions and publishing defaults, and performance controls. It remains available to the ShootFeed owner for private testing. Do not suggest TikTok setup or controls to beta/public users. Existing owner connections and saved settings are preserved. Public access will be enabled only after approval.

Mentions Only removes written prose and keeps the active platform’s included names, handles and credits. It changes that caption only, without an AI request. If no mentions or credits are included, add them first. Apply to All remains a separate action. For the owner’s selected YouTube video, YouTube’s caption starts from Instagram and participates in Apply to All; see the YouTube guide for title defaults.

Undo and Redo stay beside the character count in fixed positions, muted and disabled when unavailable. Undo restores the previous caption/title draft; Redo reapplies an undone step. Typing, generation and Mentions Only are included. Histories remain available while switching platforms or collapsing their panels and reset after saving or opening another post. If every edited field returns to its saved value, the detail editor clears the unsaved state and the inline Save Changes button is disabled. Other unsaved edits still need to be saved or discarded.

YouTube’s caption (its video description) is optional. In the detail editor, use the switch beside YouTube in Captions to include or omit it. Turning it off keeps the saved text for later and uploads an empty description. The required Title remains editable, and an empty title still shows Needs Review and blocks approval even when the caption is off. Title generation remains available while the caption is off. This switch uses the existing per-post caption preferences and does not change selected platforms or other captions. See [YouTube’s upload reference](https://developers.google.com/youtube/v3/docs/videos/insert#errors) and [required upload interface](https://developers.google.com/youtube/terms/required-minimum-functionality#uploading-videos).

Video AI captions use intelligible speech together with representative frames for clips up to ten minutes. Speech context is cached privately per source revision, so regeneration reuses it. Silent or unclear audio uses visual context without inventing dialogue; failed analysis does not prevent drafting. If another request is already preparing speech, wait a moment and retry. AI captions do not approve or publish the post.

Pinterest has an Auto-generate descriptions switch alongside the other platforms. Its caption is Pinterest's description, separate from its optional short title. In Shared caption mode it uses the universal caption pattern; in Individual mode it has its own pattern. Default Pinterest title (optional) accepts the same Smart Text chips as the other title patterns. Leave it blank to suggest a title from the post's named models, using the same name selection and formatting as Reddit. Without a named model, the suggestion uses a short caption line. Custom Pinterest title patterns keep their own values; suggestions stay within 100 characters and existing saved titles are not rewritten when this default changes. Connect Pinterest accounts in Settings → Profiles, then select them with the normal account buttons in either editor. Under Captions → Pinterest, choose the board and review the title, caption and destination link together. With several selected Pinterest accounts, each has its own grouped fields. There is no extra account dropdown. In the inline editor, use the left/right caption arrows to reach Pinterest in profile order. Its board, title, caption and link appear only on that platform page, without a nested Pinterest dropdown. Switching pages preserves the unsaved fields while the editor remains open.

Pinterest has the same per-post caption switch in the detail editor. Turning it off keeps the saved text but publishes an empty description. Its optional title, board and destination link stay independent. Saved fields load separately from the live board list, so you can review and edit copy while boards load. A board retry preserves unsaved text. Pinterest uses the same caption style dropdown, Regenerate and Apply to All controls in both editors. The selected style is used for the preview before saving. Apply to All copies the caption between selected platforms (and selected Pinterest destinations), without copying Pinterest boards, titles or links. Both editors use the existing Undo / Redo controls. Pinterest orders its fields as board, title, destination link, then caption. Across the editors, the caption count and Undo / Redo sit directly below the caption input and above the style/action row. Reddit shows the 40,000-character limit; YouTube retains its byte count. Closing and reopening the Pinterest accordion keeps the current fields and does not restart the board request. Pinterest and Reddit show an account heading only when multiple accounts are selected; a single account does not add a redundant label.

Caption platform rows, individual patterns and default title cards follow the selected profile's saved account order. For All Profiles, the currently open profile supplies the order, with other platforms afterward. Ordering does not change which platforms are available.

On an editable Pinterest draft, Generate / Regenerate uses the profile's existing photo or video caption style and prepares title and caption for review. It does not save or publish. In the inline editor, use the same Cancel and Save Changes footer as the other platforms. Save Changes saves edited captions across the platform pages together; Cancel discards unsaved edits. If a save fails, the editor stays open with the draft and error so you can correct it. On the detail page, Pinterest participates in the regular Save Changes / Discard Changes footer and the existing unsaved-change prompt. There is no separate Pinterest save button. Neither save action publishes the post. Saved per-post copy, including deliberately blank fields, keeps its own values when defaults change. Pinterest descriptions allow up to 800 characters; generation reserves room for the pattern and credits and reports a length problem instead of silently removing credits.

In the owner's Reddit editor, choose a community with the standard dropdown. Each community appears once, with a small NSFW badge where applicable. The caption accordion's status dot indicates whether attention is needed. Unavailable choices remain disabled, with their eligibility reason available on hover. Manage the community list in Settings → Communities; the post editor does not add new communities. In the detail editor, use + to add another community dropdown and − to remove that row. Every selected community stays visible, up to three per Reddit account. A compact 1/3 count beside Community updates as rows are added or removed. Each dropdown excludes communities already selected elsewhere on the post. Bot’s choice lets ShootFeed choose that slot when the draft is saved; manual selections stay selected. Community and Title appear above Caption. The Reddit status dot checks its destinations and titles; missing Tags & credits has its own separate warning. The first community receives the original post. Additional communities use Reddit's repost flow and are staggered two hours apart. These choices use the usual Save Changes and Discard Changes controls; adding a row does not publish anything. The old Choose top two again button is replaced by these explicit destination rows.

When more than one account from the same platform is selected for a post, the caption editor shows All followed by the selected account chips, in profile order. The row scrolls horizontally when needed. All starts selected and edits the shared caption. Choose an account to customize just its caption; that version survives shared-caption edits, saving and reopening the post. An intentionally empty account caption stays empty. Accounts without a custom version inherit the shared caption. One selected account does not need an account-chip row.

Regenerate and Undo / Redo act on the current caption scope. Apply to All explicitly copies the current caption across all selected platforms and accounts, replacing their custom caption versions. Captions for deselected accounts remain saved for later. Board, community, title, link and publishing choices remain separate from this caption-copy action. Use the usual Save Changes and Cancel controls; the account editor has no separate save workflow. Caption limits apply to every selected account version before approval and publication.

Both editors label every caption input Caption, including platforms with only one text field. Field labels use muted sentence case. TikTok uses Title for its photo-post title, matching Reddit, Pinterest and YouTube. Fields required for publishing have an asterisk. The inline editor uses these markers without an extra required-fields note; the detail editor includes a **\* Required** key. Optional titles, captions and destination links have no extra optional label. Drafts can still be saved before required publishing details are complete.

The account chips use the same neutral fills, borders and hover states as the Performance platform chips in each theme. Their compact height matches the platform chips in Post details. Selected blue chips keep white labels and icons.