## Summary
Reworks the desktop (Tauri) app's Create Image / Create Video pages to
match the webapp,
extracts the webapp's generation gallery into a shared library used by
both apps, embeds the
model selector into every prompt box, adds a folders sidebar + marquee
selection to the gallery
modal, restyles the login/pricing/credits modals to the webapp look, and
turns the Apps page
into the default landing tab.
### Shared generation-list lib
- Extracted the webapp's generation feed into
`@storyteller/ui-generation-list`: list + masonry
grid views, pending/failed/completed rows and cards, thumbnail handling,
gallery/prompt data
hooks, and the shared grid/list view store + toggle. Host-specific
actions (recreate,
make-video, share, download, toasts) inject via render props — webapp
markup unchanged, its
old files are now thin wrappers.
### Desktop create pages
- Prompt box fixed at the bottom, `max-w-5xl` like the webapp; pages
renamed to
"Create Image" / "Create Video".
- Generation feed: in-progress/failed from the Tauri task queue, history
from the library API,
rendered as grid or list (toggle in the TopBar next to credits). Typed
Tauri completion events
push finished media in instantly; the poll reconciles and dedupes.
- Row/card quick actions (recreate, make video, copy link, download)
backed by a shared
`desktopMediaActions` module extracted from the TopBar gallery wiring.
- Hero + background act as a splash that fades when the feed loads;
bottom gradient behind the
prompt box. Batch generations show "Generating N images" on the single
pending chip.
- Video prompt box now expands vertically as far as the image prompt box
(viewport-relative).
### Embedded model selector
- `ClassyModelSelector` gained `variant="embedded"`: a compact pill that
opens the webapp-style
rich-list popover while keeping the desktop's provider selection.
Embedded on all five
prompt-input pages; floating selectors removed; legacy "10 sec." badges
dropped.
### Gallery modal
- Left sidebar lists the user's folders (drag-and-drop, right-click
management, new-folder
button) and shows on the Folders tab too, highlighting the active
folder's root branch.
- Marquee (drag-rectangle) selection in view mode, with edge
auto-scroll; starts only on blank
background so it doesn't interfere with drag-to-folder/canvas.
### Restyled modals (desktop)
- Login/signup: webapp 2-column layout (form left, Vimeo showcase
right), no Google button
(Google auth isn't wired for Tauri), `artcraft-icon` logo, eye toggles.
The dismiss button is
removed so login is required; the post-auth Discord + Success onboarding
steps are kept, as is
the existing Tauri auth flow.
- Pricing + credits: match the webapp visually (unified monochrome plan
cards with
Most Popular / Best Value highlight frames; `#161618` credit-pack cards
with Popular badge).
All billing stays on the existing Tauri `invoke` calls — no billing
logic changed.
### Apps page / navigation
- The Apps page is now the first tab and the default landing tab, with a
Home icon. Restyled to
the webapp home: "What will you craft today?" hero, Create / Edit
sections, gradient-hover
cards sized to their content, full-width scrollbar. The TopBar
quick-menu header reads
"Create" instead of "Generate".
### Fixes
- Video Editor no longer crashes on the desktop app (`Tooltip must be
used within
TooltipProvider`): the editor shell now mounts its own Radix
`TooltipProvider`. The webapp got
one app-wide via `SidebarProvider`; the Tauri app had none, so panel
tooltips threw.
### Misc
- `getCreatorIconPathForModelId()` added to `@storyteller/model-list`
(model id → creator icon
path, desktop/webapp aware).
- **New folder keeps selection:** "Add to folder > Create new folder"
now drops
the selected items into the folder on create instead of leaving it empty
(webapp library + desktop gallery modal)
- **Marquee dead below the fold (main bug):** flex-column scroll parent
shrank
the page root to one viewport; presses lower down missed the handler.
Fixed
with `shrink-0` so the root spans the full scroll height
- **Edge auto-scroll:** holding a drag near the top/bottom edge scrolls
the list
so you can keep selecting past the viewport in one drag (anchor
compensated for
scroll; virtualized tiles stay selected)
- **No blue highlight:** `select-none` on the page (inputs opted back
in) blocks
accidental text/image highlighting and stops it competing with the drag
- **Rectangle no longer covers the top bar:** visible marquee box
clamped to the
gallery viewport; selection hit-testing stays unclamped
The desktop video promptbox built the GenerateVideoRequest without ever
setting `resolution`, so generation always omitted it and the backend
fell back to the model's default (720p) — and billed for it. The cost
preview reads the same store value and sends it, so the preview was
correct (e.g. 117 for 480p) while the charge stayed at the 720p price
(240). Populate request.resolution in buildRequest, mapping the store's
legacy display string ("480p"/"720p"/"1080p") to CommonResolution,
guarded on the model exposing resolutionOptions.
Also fix useImageCostEstimate (webapp) omitting `quality` from its
effect deps, which meant the image cost preview didn't refresh when only
the quality option changed.
…perf
- Add desktop-style marquee drag selection on the webapp library: drag
on blank space to select covered tiles, click blank space to deselect
all
- Multi-select: clicking anywhere on a tile toggles it (fixes a
double-fire that cancelled the toggle so only the checkbox worked);
selection persists across tabs and folder navigation
- Gallery item menus show "View image" / "View video" / "View 3D" above
Add to Folder; menu panels size to content so labels no longer wrap
- Mobile/touch: no tile dragging or marquee (finger drags scroll),
checkbox and menu buttons always visible (no hover on touch), and the
Add to Folder submenu expands inline as an accordion instead of
overflowing the screen
- Select-mode gallery modal: folder cards hide the options menu and
ignore right-click (picker is browse-only)
- Perf: bulk selection moved to a module store with per-tile
subscriptions so selection changes re-render only flipped tiles, never
the page; marquee rectangle is positioned imperatively with cached tile
rects and commits only when coverage changes; webapp date groups
virtualized via the shared LazyDateGroup for smooth scrolling on large
libraries
- Bulk selection bar extracted to its own component, fixed to the
viewport bottom and centered within the content area
- Redesigned the gallery modal with a sidebar replacing the old filter
dropdown - filters now show as a list with item counts
- Added a Folders system - create, rename, and delete folders to
organize assets
- Users can drag items onto folders in the sidebar, including
multi-selected items
- Added "Add to Folder" option in the item right-click menu with a
folder picker submenu
- Added "Add to folder" button in the bulk selection bar with a folder
picker popup
- Clicking a folder in the sidebar opens a folder view with its own
content area and empty state
- Folder names are editable - click the title in the header or
right-click a folder in the sidebar to rename
- Videos are now draggable - previously only images and 3D objects could
be dragged
- Folder data is mocked locally for now - ready for API integration
### Outage handling
- Show an error toast when the OmniGen service is unavailable (5xx)
instead of failing silently.
- Distinct messages for the two failure points: couldn't load the model
list vs. couldn't enqueue a generation.
- Model-fetch toast fires once per failed attempt (singleton fetch +
message dedup); generation toast is gated to 5xx so content-policy/4xx
rejections still just show the failed card and 402 still opens the
credits modal.
### Prompt box
- Fix prompt box overflowing past the viewport at some desktop widths
(Generate button hidden) - removed `w-full` so it sizes to the content
area beside the sidebar (create-image + create-video).
### Recreate action
- Add **Recreate** to in-progress jobs in the create-page gallery (grid
+ list) and the task queue - previously only on completed/failed.
- Task queue had no Recreate at all; added it to all three states
(in-progress/completed/failed) for consistency.
- Recreate on the in-progress grid card sits bottom-right inline with
the prompt; in-progress list row reveals it on hover (matching completed
rows).
- Extract shared `useRecreateFromPromptToken` hook; refactor
FailedCard/FailedRow onto it (DRY).
### Model display names
- Consolidate to a single source of truth: webapp `getModelDisplayName`
now delegates to `@storyteller/model-list` (removed the duplicate local
map). Gallery, task queue, and lightbox now agree.
- Fix missing names: `gpt_image_1` (gallery) and Midjourney in the
lightbox - added `midjourney_7` / `midjourney_7_niji` / `midjourney_8`
to the model-list name + creator-icon maps; aligned `veo_2` → "Google
Veo 2".
### Sticky prompt settings across model switches
- Prompt-box settings no longer reset when switching models - the user's
choice persists and is resolved against the new model (kept if
supported, else falls back for display/generation only, without losing
the preference).
- Covers aspect ratio, count, resolution, quality (image) and aspect
ratio, resolution, count, duration, sound (video).
- Add `lib/resolve-model-setting.ts` (`resolveModelOption` /
`resolveModelCount`); `handleModelChange` now only switches the model (+
capability-bound input mode on video).
The frontend migrated to pnpm workspaces in #1606, which introduced
"workspace:*" dependencies. npm cannot resolve that protocol, so the
macOS and Windows publish workflows fail at "npm ci" with
EUNSUPPORTEDPROTOCOL, breaking production release builds on main.
Add a pinned pnpm setup step and replace the npm install block with
"pnpm install --frozen-lockfile". Verified the frozen install resolves
cleanly against the committed pnpm-lock.yaml.
---------
Co-authored-by: bflatastic <bombayapirat@gmail.com>
- Renamed the "Supercharge Credits" button to **Use on Web** (globe
icon) in the hero and bottom CTA. Routes subscribers to the webapp
homepage, everyone else to `/pricing`.
- Removed the "Create image / Create video" links from both sections.
- Split Discord + GitHub into a separate "Open source &
community-driven" CTA (subtle container) above the "Ready to craft?"
CTA.
MentionTextarea's DOM-sync effect rewrote innerHTML under the live caret
whenever buildHTML's identity changed, detaching the cursor on iOS
Safari mid-typing. Skip the rewrite when focused and the DOM already
matches value, and memoize colorMap in MobilePromptForm to stop the
per-render churn.
After Google sign-in, brand-new accounts now land on /welcome instead of
the set-password page; returning users go to their normal destination.
"New user" is detected via the SSO response's
username_not_yet_customized (forced to false by the backend for
returning Google logins, so repeat logins never re-trigger it). A
per-user-token localStorage marker guards against the rare
email→Google-link edge case so welcome shows at most once per account.
- session.ts: add consumeNewUserWelcome(isNewUser) helper
- login / signup-form / signup-cta-modal: route via the helper
On mobile Safari, opening any overlay (mobile menu, signup modal) over the
create pages froze the UI for 5-10s. Cause: the fixed prompt box uses the
`.glass` class (`backdrop-filter: blur`), and iOS Safari re-evaluates a
backdrop-filter against the newly composited overlay content, hanging the
main thread. Desktop was unaffected.
- Disable every `backdrop-filter` on mobile (<= 639px) via a universal reset.
- Make `.glass` fully opaque on mobile so dropping the blur is invisible.
- Revert the earlier Google Sign-In deferral attempt in GoogleLoginButton
(it didn't address the real cause).
**Problem:** On mobile Safari, tapping "Sign up to create" on the create pages froze for 5–10s before the signup modal appeared. Desktop was fine.
**Cause:** The modal mounts Google's GSI sign-in widget (`<GoogleLogin>`), which boots a cross-site iframe to `accounts.google.com`. Mobile Safari's tracking prevention (ITP) throttles that cross-site init for several seconds, and because the widget mounted synchronously when the modal opened, the modal's first paint was blocked behind it. The login/signup pages didn't show this because they mount the widget during a route change, hiding the cost.
**Fix:** Defer mounting the real GSI iframe until after first paint (`useEffect` + `requestAnimationFrame`). The custom "Continue with Google" button shows instantly; the actual Google button hydrates a frame later, off the critical path. The container has a fixed height, so there's no layout shift.
- Centralized in `GoogleLoginButton`, so the login/signup pages benefit too.
- Tradeoff: the Google button becomes clickable ~1 frame later (imperceptible).
- No behavior change on desktop.
bflatastic