AI short-drama production workspace

YanDong AI
Let a director run the camera

Twenty directors' camera language, uninterrupted AI execution, and your complete toolchain — from script to finished film inside one application.

20director styles

54system art styles

MCPAI tool access

0cost for local rendering

AI wired in, ideas become actions

Connects AI to the production software, assisting with natural language to create projects, organise scripts, submit generation jobs and read results.

yandong-ai — production

› Turn chapter three of this novel into a vertical short drama in Wong Kar-wai's style, storyboard episode one first

Create project ✓ #207 · 9:16 · Wong Kar-wai

Split episodes ✓ 6 episodes

Extract characters and scenes ✓ 8 characters · 5 scenes · 3 props

Generate storyboard ✓ Episode 1 · 4-panel grid

Submit dubbing ✓ Task running · quote retrieved

All six steps initiated by AI. Zero clicks in the software.

MCPpublic tool interfaces

11production stages covered

4connected AI hosts

2directions: read software / write back

Twenty directors on set, no more generic camera language

Each director ships a complete set of storyboard and shooting rules, plus dedicated art direction for characters, scenes, props and storyboards. Not a filter — a different person calling the shots.

All 20 directors ready Each = 1 reference image + 4 art prompts + 1 complete shooting brief

STYLE · 20 / 20

01Steven SpielbergEnter the event through the character's perception

02Alfred HitchcockOrganise the audience's suspense

03Stanley KubrickThe order and central axis of space

04Akira KurosawaBody posture and group blocking

05Martin ScorseseFollow the character's purpose

06Federico FelliniA constantly reorganising living scene

07Ingmar BergmanThe afterglow on speaking and listening faces

08Andrei TarkovskyTime flowing through materials

09Wong Kar-waiApproach, avoidance, and the unfinished

10Christopher NolanMechanics of action and physical consequence

11Denis VilleneuveScale, pressure, and the gaze

12David FincherCracks beneath a calm surface

13Nicolas Winding RefnNight, neon, and urban solitude

14Bi GanDampness, loss, and poetic long takes

15Zhang YimouBroad colour fields and ceremonial blocking

16Hou Hsiao-hsienLocked camera and events within the frame

17Park Chan-wookMoral pressure closing in

18Terrence MalickDrifting time and interior monologue

19Michael MannUrban night and professional action

20Coen BrothersDeadpan humour at the edge of control

Once a director is selected, all three storyboard flows — single-shot S, grid G, multi-reference M — adopt that director's rules automatically. Composition, colour and material preferences land on every character image, scene image and every storyboard panel. Styles without a dedicated brief fall back to the general standard.

Fits your local workflow, turns your GPU into production capacity

Drop in a ComfyUI workflow you have already tuned. The software detects nodes, converts formats, injects assets and computes frame counts; finished output lands straight in the project.

Your ComfyUIraw workflow JSON exported from the UI

Auto-adaptdetect nodes · convert · compute frames · inject assets

Project assetsoutput lands in the project with full history

VideoMiniMax H3 multi-referenceimage, text and audio references together, 30s direct output, 6 workflow variants adapted

VideoH3 image-to-video acceleratedBF16 high-quality pipeline, 10 Eros mixed Turbo variant

ImageQwen Image 2.1joins the ComfyUI image channel, node names and model paths matched automatically

AudioFish Speech S2 / Breeze TTS 2voice cloning and line-by-line dubbing, usable as reference audio fed back into local video workflows

AnyAny custom workflowreads UI-format JSON directly and names the exact node that is missing

Asset injection: reference images, prompts, voices, driver audio, duration and seed are filled in automatically
Frame math: converts to each model's legal frame count, so duration mismatches never surface
Recovery: finds the job back in the queue by its unique id — never re-burns the GPU
Dual channel: images and video are configured separately; local and cloud can be switched per job
Write-back: externally made images, audio and text are written back by asset id, never by guessed name

No per-second billing

Local rendering consumes no credits. 100 clips a day costs the same as 1.

No queueing

Cloud peaks mean waiting; local runs when you want, and keeps running offline.

Nothing leaves the machine

Script, characters and voices stay on your own computer.

From script to audiovisual assets, one connected pipeline

Script splitting, characters and scenes, storyboards, image generation, dubbing and video generation all live in one project, so you stop switching tools.

Split scriptautomatic episodes, editable format

Characters & scenescharacters, scenes and props stored

Storyboarddirector's shooting brief takes over

Imagesgrids and top views

Dubbingline-by-line dialogue and scene audio

Deliverycloud and local on equal footing

Four production modes: short drama · narration · grid-to-video · multi-reference · landscape and vertical both supported.

Pick a director, make your first film

Windows client, install and go. Your project data stays on your own machine.

Download

Fetching the latest version…