Changelog

Features
- canvas: add owner-gated videoProcessing (ffmpeg) node
- canvas: distinct image-input role field + per-image-param slot cap (Phase 1)
- canvas: Phase 2 distinct image input sub-slots (per-role handles + groups + caps)
- canvas: Kling 3.0 start/end image frame sub-slots
- canvas: add Note sticky node (x-header + resizable textarea, no connections, layout-pinned)
- canvas: show input video duration in video-gen Duration chip when model has no duration param
- canvas: videoProcessing resolution option, grain cap 15, bt709 color fix, temporal sensor noise

Fixes
- canvas: apply forwarded className in StaticDrawerBody; drop unused X imports
- canvas: break edge link when removing connected text item from Google Sheets write column
- canvas: make Tooltip-wrapped node dropdowns open on touch
- lint: post-merge lint cleanup (prettier + cognitive complexity)
- canvas: large processed-video cloud uploads + lower grain/noise cap
- canvas: scope single-play recovery to the playing node only
- canvas: persist generation history on transient callback failure
- canvas: videoProcessing node saves output to library + adds history button
- canvas: supabase-backed nodes emit durable media-proxy URL, not 1h signed URL

Features
- canvas: simplify google sheets node per-row type model + last-frame
- canvas: enable google sheets Run with inputs or outputs
- canvas: add Kling 3.0 Motion Control video-to-video model
- canvas: Google Sheets node handle cleanup, GS-source cursor gating, unified write-column reorder

Fixes
- canvas: apply gemini-omni with-video-input flat surcharge in cost estimate
- canvas: bind kie callback/retry to failing task_id to stop stale-task fail+refund
- canvas: fan out google sheets text output per read column
- canvas: gemini-omni-video connected video defaults to video-only
- canvas: Google Sheets per-column connected media outputs feed and remove independently
- canvas: resolve Google Sheets connected media display for last-frame, audio, and video thumbnail

Features
- canvas: replace node param-title labels with lucide icons
- canvas: add Topaz video-upscale model (kie topaz sub-provider)
- canvas: float node run-error as a card above the node
- canvas: add gemini-omni-video video input + conditional with-video pricing
- canvas: collapse node upload/import into + popover and replace history dropdown with selection modal

Fixes
- canvas: bill seedance v2v on uploaded/imported source video duration at quote time
- canvas: count connected uploaded/imported source video duration in v2v cost estimate
- canvas/video-gen: make v2v input-duration estimate reliable + detect imported video
- canvas: enforce google-sheets connection matrix spec + media-kind type propagation
- canvas: refund credits only on provider POST failure or kie.ai generation failure
- security: harden codebase from full-surface security audit
- canvas: parallelize node multi-file upload/import instead of sequential
- canvas: harden single-play & run-all mid-run recovery hydration
- canvas: enforce node-items-types-matrix across all 9 node types
- canvas: drop unused useMemo import after apex-merge conflict resolution
- canvas: pass Google Sheets per-kind media output downstream + fix text-removal cascade
- canvas: total cost badge mirrors per-node video pricing (effective variant + input seconds)
- canvas: total cost badge sums published per-node estimates instead of recomputing

Features
- i18n: add Russian (ru) locale end-to-end + geo-IP detection
- canvas: reasoning/thinking effort for text-gen models (openai, xai, anthropic)
- canvas: redesign node UI/UX - compact|large view, glass settings drawer, inside-node preview
- canvas: settings drawer collapsed band -> centered glass pill
- canvas: move node-info i button into node picker drawer
- canvas: text-node copy/count/clear header, fix node resize animation, drawer-aware layout
- i18n: add Russian (ru) locale end-to-end + geo-IP detection
- canvas: reasoning/thinking effort for text-gen models (openai, xai, anthropic)
- canvas: redesign node UI/UX - compact|large view, glass settings drawer, inside-node preview
- canvas: settings drawer collapsed band -> centered glass pill
- canvas: move node-info i button into node picker drawer
- canvas: text-node copy/count/clear header, fix node resize animation, drawer-aware layout

Fixes
- i18n: add missing largeView (and compactView) keys to ru.json
- canvas: node UI tweaks - textbox centering, pulse radius, preview close animation, text-gen download removal
- canvas: bill seedance v2v on source video duration
- canvas: single-play creates a running run-log so the blue project dot + mid-run recovery work

Features
- canvas: gray out run-all when a node has a missing required field
- canvas/google-sheets: collapse output-kind tiles into hover-expand row
- canvas/google-sheets: multi-condition row filter (AND/OR), column-hide selector, wider node, refresh + border handles
- canvas/google-sheets: write-column rework - shared media indicators, multi-text append, single-media, formula columns
- canvas/google-sheets: write-column manual input + shared media indicators

Fixes
- canvas/google-sheets: self-heal orphaned connectionId in node account selector
- canvas/buffer: gate play button on title + keep Title link across reload
- canvas/gen-nodes: preserve uploaded/imported media on model switch
- canvas/gen-nodes: preserve uploaded/imported media across model switch (key-remap)

Features
- canvas/google-sheets: redesign node UI + wire text-source end to end
- canvas/google-sheets: full media participant (source + write-back target)
- canvas/google-sheets: redesign node UI - per-row handles + Input|Output column rows

Fixes
- buffer: reduce API call amplification driving rate limits
- canvas/google-sheets: connection rules + audio item display + multi-slot edge removal
- api/google-sheets: proactive OAuth token auto-renewal

Features
- canvas: text-gen temperature as header popover-slider, bounds 0.4-1.6
- buffer: OAuth app-client multi-account (replace BUFFER_API_KEY)
- canvas/buffer: redesign buffer node UI + split connection handles + YouTube Short
- canvas: internet-browsing (web search) for text-gen models + planet toggle
- canvas/buffer: unified text-input lists, multi-title, animated layout + handle/bracket fixes

Fixes
- canvas: allow provider media host in CSP connect-src so node preview downloads work in prod
- canvas: media node video last-frame downstream + buffer recognition + playable preview
- buffer: multi-account connect/switch, node auto-pick, settings header + confirm-disconnect
- buffer: apply #Short to YouTube channels only, server-side per-channel
- canvas/buffer: batch runner resolves media-node upstreams via effectiveSourceType
- buffer: repair token refresh, OAuth grant expiry, and multi-account channel selection

Features
- credits: round USD->credits up so stored credits never fall below provider price
- models: enforce per-model prompt character limits across image/video seeders
- canvas: text-gen cost estimate counts upstream input tokens
- canvas: gray out single-play button when a node has an awaiting input
- canvas: dedup buffer node title + collapsible per-platform groups
- canvas: add temperature parameter to grok-4.3 text-gen model

Fixes
- models: enforce eleven_v3 prompt char limit (maxLength + maxTotalPromptLength = 3000)
- models: drop unsupported grok-imagine-video-1.5 text-to-video variant (xAI is i2v-only); deactivate prod row
- analytics: prod-only PostHog capture, internal opt-out, signup funnel + noise cleanup
- canvas: clear stuck isRunning lock when a generation node fails during recovery
- billing: audit + harden Stripe subscription/payment/gating/cancellation flow
- credits: close TOCTOU double-reset race on subscription_reset

Features
- models: add xAI grok-imagine-video-1.5 + 1-15s duration for grok video
- landing: show Grok Imagine Video 1.5 in hero marquee + doc the step
- canvas: add per-model prompt character limits with real-time UI feedback
- analytics: expand PostHog capture + wire server-side error tracking
- onboarding: phase-2 first-time pulse hints (new-project button + node drawer)
- pricing: grok video price scales with resolution (480p/720p)

Fixes
- canvas: clean grok video 1.5 display name + collapse crowded duration ticks
- canvas: gate single-node play button on visible required fields for text/audio nodes

Features
- models: add Claude Opus 4.8 text-gen model

Fixes
- canvas: allow upstream media node to connect to text-gen node
- canvas: openai text-gen no longer fails for all gpt-5.x models
- canvas: correct image-gen node input consumption across providers
- canvas: drop reasoning param from openai text-gen entirely
- canvas: consume + post video-node inputs by model input-type across KIE/XAI
- canvas: show downloading bar the instant the generating bar disappears
- canvas: honor connectedInputOrder in batch + API upstream-text resolvers

Six competitor comparison pages (/vs/kling, /vs/runway, /vs/pika, /vs/submagic, /vs/opus-clip, /vs/captions) and five audience landing pages (/for/creators, /for/agencies, /for/faceless-youtube, /for/short-form-editors, /for/ecommerce) are now live.

A press kit page (/press) provides the company boilerplate, founding story, brand assets, and brand colors.

This changelog page (/changelog) indexes product updates with JSON-LD structured data.

All new pages include JSON-LD structured data (CollectionPage, BlogPosting, WebPage, BreadcrumbList) and hreflang alternates for en, fr, and zh-CN.

Flowctory is now open to the public as a beta product.

The platform ships a node-based visual canvas where users chain image generation, video generation, and video stitching steps. Generated clips are assembled by an FFmpeg pipeline and optionally published directly to TikTok.

Account plans (Free, Plus, Pro, Max) are live. Paid plans include a 14-day free trial.

The interface is available in English, French, and Simplified Chinese.

Support reaches users at contact@flowctory.com.