End-to-end workflow
Stories, asset libraries, linear timeline, advanced prep, and FFmpeg export in one desktop app.
YSK product · AI short-drama studio · MIT · v1.3+
InstantDrama Magician
AI professional short-drama desktop tool. From idea to finished film:
story → cast / costumes / scenes / props / actions → linear timeline → AI storyboard & video → FFmpeg export.
Electron desktop + browser remote + CLI instant-drama (~157 channels).
End-to-endFull pipeline to final film
ActionsMulti-panel action guides
MediaGenUnified polish / gen shell
FFmpegExport, xfade, subtitles
From idea to finished film
Story, assets, timeline, and export in one pipeline — scan the flow and capability matrix below.
- StoryAI style note, beats, cast binding
- Asset librariesCharacters / costumes / scenes / props / actions
- TimelineLayout clips, beat screenplay, snap
- GenerateStoryboard stills → video queue
- ExportFFmpeg, TTS, subs, xfade
Actions motion guides
Multi-panel boards attach to cast, beats, and timeline clips — inject pacing and camera intent.
MediaGen shell
Libraries and timeline refine share materials → multi-image vision polish → image / video.
Remote + CLI
Browser remote on the same data; ~157 channels for scripts, CI, and agents.
| Area | What you can do |
|---|---|
| Stories | Multi-story management, cover AI, style bible, beats, cast binding, .idm.zip backup |
| Characters | Global cast, soul.md / SoulMD Hub, multi-angle sheets, identity lock, vision fill |
| Costumes | Wardrobe library, try-on dual-write to character + costume galleries |
| Scenes | Copy, plates / looks / atmosphere, scene gallery, vision AI |
| Props | Descriptions, master prompt, plate variants, timeline binding |
| Actions | Multi-panel boards (2–6), cast refs, inject pacing into generation |
| MediaGen | Materials → multi-image vision polish → image / video |
| Gallery UI | EntityGalleryPanel: preview, cover, identity multi-select, intro video |
| Timeline | Linear layout, snap/pack, 6s/10s clips, dialogue & camera tags |
| Advanced prep | Cast lock → storyboard stills → video; per-cell refine |
| Audio / subs | TTS mix, burn-in subs, xfade / ducking, aspect-aware export |
| Settings | LLM / image / video providers, FFmpeg, web server, auto-update |
| CLI / remote | ~157 channels; in-app web or standalone instant-drama server |
| i18n / updates | 10 UI languages; packaged GitHub Releases auto-update |
Creative asset libraries
Reusable global assets — build once, share across stories.
Stories
- Create / edit / delete independent short-drama projects; status, cover, sort
- Cover zoom / regenerate / save as; import/export story
.idm.zip - Basics: cover, AI quick create (style note / beats), title, status, art style, style bible
- Cast: link characters, scenes, props, actions (search + in/out filters)
- Script beats: multi-select assets per beat including actions; beat screenplay text
Characters
- Global cast library; filters for gender, art style, image, soul, language
- Profile: name, description, age, gender, language, voice; AI fill from concept/draft/soul or still-only vision
- References: multi-angle bible, body/base/costume pipeline, external refs, Lock identity, intro video
- Costume tab binding; SoulMD Hub index + soul.md import
Costumes
- Global wardrobe linked to 0…N characters
- Vision fill from reference stills only; multi-image gallery, cover, intro video
- Identity multi-select thumbs; try-on with character refs
- Accept try-on draft dual-writes to character and costume galleries
Scenes
- Description and script fields; plates, looks, atmosphere
- Scene gallery and variants; vision AI fill from selected/cover still
Props
- Name, description, master prompt, plate variants
- Timeline clip binding; vision AI fill from reference stills
Actions (motion guides)
- Global action library for body / blocking / staging cues
- Multi-panel boards: 2–6 panels (strip or 2×2 / 2×3); panel 1 = first beat, panel N = last
- Art style, external refs, cast refs from characters/costumes/scenes/props
- Vision AI fill; accumulate panels, reorder, cover
- Attach to story cast, beats, and timeline clips; inject pacing / intent / camera notes; optional i2v reference
MediaGen shell
- Unified: materials → multi-image vision polish → image / video
mediaGen:extract·polish·generateImage- Shared by library pages and timeline refine (timeline-still / timeline-clip)
Gallery UI
- Shared EntityGalleryPanel: large preview, zoom / save-as / cover / remove / intro video
- Thumbnail row: preview vs identity multi-select
- Labels and errors localized
Timeline & export
Production desk: layout, advanced prep, audio/subs, and FFmpeg final export.
Timeline
- Play / Undo / Redo; batch Generate; Export + history
- Duration, ready count; AI clips 6s or 10s only
- Zoom, timeline snap, snap grid, pack clips
- Clip editor: bind characters / scenes / props / actions, duration, beat screenplay tags
- Per clip generate / regenerate / continue; cancel; retry-failed-only
- Export options: TTS, burn-in subs, xfade, ducking, aspect ratio
Advanced prep
- From Timeline → Advanced
- Cast lock on-screen looks
- Storyboard stills: tail-frame continuity, prev keyframe base, multi-ref polish, backfill missing, status labels
- Video queue when stills ready (skip existing optional)
- Per-cell refine still or video via MediaGen
Audio / subtitles
- Optional TTS mix and burn-in dialogue subtitles
- xfade transitions, BGM ducking, aspect-aware export
Tools & settings
Activity log, providers, i18n, and auto-update.
Activity log
- Local activity.jsonl-style events: generation, export, update, support reports
- Debug API / pipeline issues
Settings
- LLM: OpenAI-compatible; default Grok Gateway (:3847); OpenAI / Custom / Kimi…
- Image: shared or independent base URL / key / model (e.g. Seedream)
- Video: auto / http / stub; Seedance, Grok
/v1/videos; 6/10s; poll & timeout - Diagnostics, FFmpeg path, web server, auto-update, support report (keys redacted), legal AUP
i18n & auto-update
- zh-HK, zh-CN, en, es, hi, ar (RTL), pt-BR, fr, ja, ru
- Packaged electron-updater against GitHub Releases
Honest limits: final look depends on your models and prompts. This tool owns workflow, continuity, and export — not guaranteed cinema-grade auto film. Store signing / notarize needs your certificates.
CLI & web remote
Headless scripts, agent integration, or browser control of the same project data.
CLI instant-drama
npm i -g instant-drama-magician- local vs remote (
/api/invoke+ Bearer) - ~157 channels covering desktop IPC; OpenClaw / Hermes agents
- Domains: stories, characters, costumes, mediaGen, timeline, videoPrep, actions, generation, server, tools schema…
Web remote & self-host
- A: in-app web server (Settings) — same userData as desktop
- B: standalone
instant-drama serverwith data dir + auth token
Screenshots & install
Desktop UI at a glance, plus install steps and recommended workflow.
01
Story management
Multi-project list, covers, status, asset counts including actions, search, import/export backups.
02
Story editor (Basics)
Cover, AI style note / beats, style bible, external refs, identity options.
03
Character library
Global cast, multi-image sheets, filters; Grok CLI status in sidebar.
04
Character references
Identity / body / costume pipeline, Character bible, Lock identity, intro video.
05
Timeline desk
Clip layout, snap, Generate / Export, Advanced entry.
06
Advanced prep
Cast lock → storyboard stills (continuity) → video queue.
Recommended flow
- Settings → API key → Test Chat
- Stories → create / AI style note + beats
- Characters → multi-angle sheet → lock identity (or still-only vision fill)
- Scenes / Props / Costumes / Actions assets
- Story Cast → link assets including actions
- Script beats → bind assets per beat
- Timeline → layout clips, beat screenplay
- Advanced prep → stills (continuity) → video
- Export (optional TTS / subs)
npm install -g instant-drama-magician
instant-drama doctor --json
instant-drama --local stories list --json
instant-drama channels list --json # ~157
npm install && npx prisma db push && npm run devGet started with InstantDrama Magician
MIT open source · YSK Limited. Turn short-drama ideas into an editable, exportable, iterable workflow.