Release history
Get 1.1.0
Known issues
Every build, and what changed in it.
Five builds since August. Fixes and the things still broken sit next to the new features, because a history that only lists wins is an advertisement.
Last updated 6 Nov 2026 · 1.1.0 is current
Up next
1.1.2
In progress
Three targeted tool upgrades & fixes currently on the workbench
SC-1036Translate & Subtitle (TranslateScreen / ExportDialog): Cluster-aware safe-area line wrapping for Thai and Khmer burned-in subtitles so long lines never clip on 9:16 vertical or 16:9 reels.
SC-1038Batch Queue Cost Estimator (BatchScreen / ManageQueueDialog): Pre-run cost breakdown before RUN NOW adds a configurable transient-retry budget range alongside base API generation cost.
SC-1040Multi-Monitor Workspace Persistence (WinformScreen / VideoPlayerDialog): Save and restore window bounds, maximized state, and DPI scaling on secondary and tertiary displays across restarts.
Planned, not promised. A release date is published only when the build passes full regression and VirusTotal verification.
Follow progress on known issues
39 AI services, 80+ models, Hybrid GPU + Cloud Vision QC, Node Workflows & 12 languages
The major studio release connecting 39 built-in AI services (80+ video, image, voice, speech, translation, lip-sync, and upscale models) plus custom REST/OpenAI/fal/Replicate endpoints, local GPU + Cloud Vision quality control, visual node workflows, NLE timeline export, and a native 3-step Setup Wizard. 10 new3 fixed1 changed
New
39 Built-In AI Services, 80+ Models & Custom Provider Attachment (
BatchScreen / AddProviderDialog) — Connect 13 video services (Kling 4.0 & Omni O3, Google Vertex AI Veo 3.1 / Fast / Lite, Gemini Omni 1.1 Flash up to 40s, Runway Gen-4.5 & Aleph 2.0, Luma Ray 3.2 HDR/EXR, MiniMax H3, Seedance 2.5 with up to 30 references, Alibaba Wan 3.0 & HappyHorse 1.1, xAI Grok Imagine Video 1.5, Vidu Q4, Pika 2.5, Lightricks LTX-2.5 Pro 4K, plus Sora 2 Pro via fal.ai), 4 multi-model gateways (fal.ai, WaveSpeed, Replicate, OpenRouter), or attach any custom REST/OpenAI/fal/Replicate endpoint.
New
Expanded Image Queue & Character-Lock Generators (
ImageQueueScreen / ImageLibraryDialog) — Generate concept art, keyframes, and character turnarounds across 12 image services and 47 models including GPT Image 2 & DALL·E 3 HD, Nano Banana 2 & Pro (Gemini 3), Google Imagen 4 Ultra, xAI Grok Imagine 2.0, Seedream 4.5 & SeedEdit 3.0, Flux 2 Pro, Flux.1 Kontext Max/Pro (Character Lock), Kling Kolors 2.1 & Virtual Try-On, Qwen-Image 2.0, Hunyuan Image 3.0, Hidream I1/E1, Ideogram v3 Character, Recraft v3 Raw/SVG, Midjourney v7 & Niji v6, SD 3.5 Large, Leonardo Phoenix 1.0, and Luma Photon 1.
New
Hybrid Consistency Check: Local GPU + Cloud Vision AI Inspector (
ConsistencyCheckScreen / Check.cs) — Score every rendered shot (0–100%) against its locked @char, @outfit, @bg, and @prop references on your own GPU at $0 API cost using CIE76 Delta-E color distance and 64-bit dHash portrait likeness, with optional deep inspection via Gemini 2.5 Flash, Claude Sonnet 4.5, GPT-4o-mini, or OpenRouter to flag face drift, outfit mismatches, and anatomy defects.
New
Pre-Flight Continuity Rules Engine (
ContinuityRulesScreen / EditRuleDialog) — Automatically audit every queued row before it reaches a paid API endpoint: block dispatch when a required character, outfit, or background lock is missing, when two mutually exclusive locks conflict in the same scene, or when shot duration/resolution exceeds the target model's hardware cap.
New
Episode Storyboard, Scene Approvals, Lock Sweep & Multi-Project Queue (
StoryboardScreen / ApprovalsScreen) — Review shots in a visual scene-by-scene storyboard grid with inline playback, mark takes as Draft, Needs Retake, or Approved, run an automated episode-wide Lock Sweep, and queue multiple episodes sequentially via the Multi-Project Queue Sidebar (QueueProjects).
New
Visual Node Workflow Builder & Shot Version Comparison (
WorkflowScreen / VersionsDialog) — Chain multi-stage pipelines (Script Sheet → Prompt Expander → Reference Image → Video Render → Lip-Sync → 4K Upscale → NLE Export) with episode-level "Pause on first error" protection, and compare v1–v4 shot takes side by side with diffed prompts, models, seeds, and consistency scores.
New
Local Real-ESRGAN 4K Image Upscaling & Cloud Video Super-Resolution (
UpscaleDialog) — Upscale reference frames and generated images locally on your GPU (2×/3×/4× up to 4K with GFPGAN/CodeFormer Face Restore at $0 API cost), or upscale rendered video shots via cloud Topaz Video Upscale (2×–4×) and fal.ai Video Upscaler.
New
AI Lip-Sync & Expressive Voice TTS Pipeline (
LipsyncDialog / Speech.cs) — Synchronize character dialogue to video shots using Kling Lip-Sync (1080p), sync-lipsync v2 (up to 4K 4096×2160), and Hedra Character 3, paired with ElevenLabs v3 (emotional tags), ElevenLabs Flash v2.5, OpenAI Realtime Mini WebSocket TTS, and Azure Neural HD voices.
New
End-to-End Video Translation, Subtitling & Auto-Speed Dubbing (
TranslateScreen / Translate.cs) — Transcribe existing videos via local GPU Whisper large-v3, OpenAI Whisper-1, Deepgram Nova-2/3, or Gemini 2.5 Flash; translate contextually with DeepL, Google Translation v3, GPT-5.1, or Claude Sonnet 4.5; burn in styled .srt/.ass subtitles; and synthesize dubbed voice tracks with automatic time-stretch fit up to 1.38×.
New
NLE Timeline Export, Local Webhook Server, 3-Step Setup Wizard & 12 Live Languages (
ExportDialog / WebhookScreen) — Export approved episodes to Premiere Pro / DaVinci Resolve FCPXML, EDL, CSV, and concatenated MP4 reels; automate jobs via the built-in local HTTP Webhook server (EditWebhookDialog); switch across 12 UI languages without restarting; and install via the 3-step Setup Wizard (50.8 MB .exe / 45.2 MB .msi / 45.1 MB .zip) with System Readiness Preflight (FFmpeg, GPU VRAM, WebView2, disk space) and dual SHA-256 + MD5 verification (0/72 VirusTotal).
Fixed
High-DPI Dialog Layout & Vietnamese Typography Scaling — Resolved button and label clipping at 125%–150% Windows display scaling by converting Export, Lip-Sync, Setup Wizard, and API Settings into DPI-aware, resizable, and maximizable dialogs.
From tester report SC-1042
Fixed
Live API Key Hot-Reload in Active Queues (
Scheduler.cs) — Jobs paused in "Waiting" after an API key expired or ran out of credits now automatically re-evaluate credentials and resume on the next 250ms scheduler tick as soon as a replacement key is saved.
From tester report SC-1019
Fixed
Multi-Project Image Library Root Isolation (
ImageLibraryDialog) — Fixed a path-prefix collision where opening the Image Library could display generated assets from a sibling episode when two projects shared the same parent directory.
From tester report SC-1022
Changed
Version-Isolated Settings & SQLite Job Database (
Paths.cs / Db.cs) — Each major/minor release now maintains its own isolated configuration directory and SQLite state file so installing 1.1.0 never mutates or locks an episode still being rendered in 1.0.x.
Overnight queue reliability, multi-key failover & atomic crash recovery
Engineered after leaving a 46-shot episode running overnight and finding it stalled at shot 9 the next morning—introducing per-service concurrency controls, multi-account API key rotation, atomic SQLite disk persistence, and pre-flight cost estimation. 2 new2 fixed1 changed
New
Per-Service Thread Limits & Dispatch Pacing (
ManageQueueDialog / Scheduler.cs) — Configure independent concurrent worker threads per AI provider (e.g., 4 threads for Kling, 2 for Vertex AI Veo, 6 for fal.ai) plus a customizable inter-row dispatch delay so high-volume queues never trigger HTTP 429 Too Many Requests rate-limit stalls.
New
Multi-Account API Key Pool & Automatic Quota Failover (
AddAccountDialog / ApiSettingsDialog) — Register multiple API keys or service accounts under a single provider; when the active key exhausts its credit balance or hits a daily quota wall mid-batch, the scheduler seamlessly rotates to the next healthy key without dropping the running shot.
Fixed
Atomic Disk Persistence on Every State Transition (
Jobs.cs / Db.cs) — Eliminated queue loss when Windows reboots or a GPU driver resets during a 40+ shot batch; every job state change (Queued → Dispatched → Polling → Rendered / Failed) is flushed atomically to disk.
From tester report SC-1008
Fixed
Smart Error Classifier for Content Moderation vs. Transient Failures (
Net.cs) — Separated safety/moderation rejections and invalid-parameter 400 responses from transient 502/503/timeout network errors so moderation blocks halt immediately with a clear reason instead of burning your retry budget.
From tester report SC-1011
Changed
Upfront Batch Cost Breakdown Before
RUN NOW (BatchScreen / ImageQueueScreen) — The queue action bar now calculates and displays the exact estimated cost across all ticked video, image, and voice jobs before you click RUN NOW, rather than waiting for the first job to return.
Enforced Reference Sets (@char, @outfit, @bg, @prop) & Excel Script Sheet import
Character, outfit, background, and prop codes stopped being manual naming conventions and became structured project-level locks enforced automatically by the queue engine alongside spreadsheet script imports.
2 new1 fixed1 known issue
New
Structured Reference Sets & Automatic Prompt Token Injection (
ReferenceSetsScreen / EditLockDialog / Locks.cs) — Create reusable @char, @outfit, @bg, and @prop reference locks containing canonical anchor images and locked visual descriptors that automatically expand into prompts and attach reference frames at queue dispatch time.
New
Excel & TXT Script Sheet Import with Delta Re-Import (
ScriptSheetScreen / ImportDialog / Sheet.cs) — Import full episode breakdowns from .xlsx, .csv, and structured .txt files with interactive column mapping (Scene, Shot, Prompt, Dialogue, Reference Locks, Duration) and delta re-import that updates only edited rows while preserving already-rendered shots.
Fixed
Unresolved Reference Code Validation Before Dispatch (
Locks.cs) — Fixed a bug where a prompt containing a misspelled or undefined @char code was sent raw to the video API, causing the model to invent a random face; unresolved @ codes now block the row and highlight the missing lock.
From tester report SC-1003
Known issue
Row-Reorder Delta Detection in Script Sheets (
Sheet.cs) — Re-importing an Excel script sheet after reordering rows (without modifying their prompt text) could mark moved rows as changed and re-queue shots that were already rendered. Resolved in 1.0.4 via content-hash row tracking.
First public release — Unified Video & Image Queues, portable episode folders & DPAPI key vault
The foundational Windows desktop release unifying multi-provider video and image batch generation around a self-contained, drive-portable episode folder architecture and local Windows DPAPI encryption. 4 new
New
Multi-Provider Batch Video Queue & Frame-Stepping Player (
BatchScreen / VideoPlayerDialog) — Unified batch queue across 11 video generation services supporting Text-to-Video, Image-to-Video, and First/Last Frame modes with per-row model, aspect ratio (16:9, 9:16, 1:1, 21:9), duration, resolution, and seed controls.
New
Batch Image Generation Queue & Reference Picker (
ImageQueueScreen / ImageViewerDialog) — Dedicated storyboard and concept image queue across 6 image providers with low/medium/high quality tiers, keyboard arrow-key inspection (← / →), and one-click "Use as reference" routing into the Video Queue.
New
Portable Episode Folder Architecture (
NewEpisodeDialog / OpenEpisodeDialog / DuplicateEpisodeDialog) — Every episode project is stored as a self-contained directory tree with paths relative to the episode root—move an episode folder to another workstation or external SSD and every script row, reference image, and rendered take opens intact.
New
Zero-Cloud Windows DPAPI Credential Vault (
ApiSettingsDialog) — All API keys are encrypted on disk using the native Windows Data Protection API (DPAPI) bound to your local Windows user profile—keys are never uploaded to any central server and connect directly from your machine to each AI provider.
Private production alpha — Initial sequential batch runner for Kling & Google Veo
Internal studio prototype shared with four series creators to automate multi-shot episode rendering. Documented here because episode folders created in 0.9.6 still open cleanly in 1.1.0 without migration gaps. 1 new1 known issue
New
Sequential Batch Video Queue Prototype (
BatchScreen Alpha) — The first working desktop batch table dispatching sequential Text-to-Video and Image-to-Video jobs to Kling and Google Veo, automatically downloading finished MP4 clips into Episode_Scene_Shot filenames.
Known issue
Single-Threaded Dispatch Without Auto-Resume — Ran strictly one job at a time with in-memory queue state; a dropped network connection or API timeout required restarting unfinished shots in the scene manually. Addressed in
1.0.0 and 1.0.4.
Nothing is taken down.
Every build stays downloadable and keeps its own settings folder, so an episode half-rendered on 1.0.4 is never forced onto a newer version mid-project. If a release breaks something you depend on, say so in the group and go back a version in the meantime.
Every version stays downloadable, with its own settings folder.
Breaking changes are called out before the feature list, never buried in it.
Known issues are published here and on the status page, not only when asked.