Product Guide v1.0 — February 2026

AI ProductionStudio

From character creation to video production — an AI-powered end-to-end production platform. One workspace. Infinite possibilities.

5
Character Stages
3
Script Phases
7
Shot Types
10min
Clip Segments
Platform Overview

Three Production Pillars

Everything that used to require jumping between multiple tools is now unified into one seamless workflow. Character design, scriptwriting, and video generation — all in one workspace.

Core Principle

“Batch Generate → Human Selects → Auto-Advance”

At every stage, AI produces multiple candidates and a human makes the final pick. Quality decisions stay with people; repetitive work is handled by AI.

Character Studio

Generate faces, photorealistic images, and outfits with AI, then composite them into consistent characters across a five-stage pipeline.

  • Face Generation — 20 candidates per batch
  • Photorealistic Rendering
  • Outfit Generation with Style Profiles
  • Full-Body Composite
  • Detail Refinement Loop

Script Studio

Write scripts, apply speech-style transforms, translate to other languages, and auto-split into 10 min clips ready for video generation.

  • Base Script Drafting
  • Speech-Style Transform Presets
  • Multi-Language Translation
  • 10-Min Auto-Split
  • Version Control & Rollback

Video Production

Produce wide, over-the-shoulder, close-up, and landscape shots as AI-generated video — with or without dialogue — up to 10 min segments.

  • 7 Distinct Shot Types
  • Wide, OTS, ECU, Landscape
  • Dialogue & Silent Modes
  • Shot Prompt Builder
  • Batch Generation Pipeline
Core Workflow

How Every Stage Works

Every production stage follows the same three-step pattern. This consistent loop means you work the same way everywhere in the platform.

01

Batch Generate

Multiple candidates are generated simultaneously. Face images default to 20 at a time; video clips default to 6. Bulk generation is the norm — not single-item creation.

Configurable batch size: 5 / 10 / 20

02

Human-in-the-Loop Gate

Results at every stage must be reviewed and approved by a person. The system never advances to the next pipeline stage without explicit human sign-off.

Quick actions: Pick / Reject / Send to Next Step

03

Auto Rank + Auto Retry

Every generated asset receives an automatic quality score from 0 to 100. Results are sorted highest-first to minimize selection time.

Detection flags: bad hands/legs, face drift, unintended audio

Workflow State Machine

Draft
DRAFT
Running
RUNNING
Needs Review
NEEDS REVIEW
Approved
APPROVED
Next Stage
NEXT
Character Pipeline

Five-Stage Character Creation

Character creation follows a sequential pipeline. At each stage the AI generates multiple candidates, and once you select the best result, it automatically feeds into the next stage.

👤
C1

Face Generation

Enter the character traits you want (gender, age range, mood, etc.) and the AI generates 20 face candidates in one batch. Results are ranked by quality score.

Batch: 20 imagesApproval Required
🎨
C2

Photorealistic Rendering

Your selected face becomes the input for photorealistic conversion. The AI preserves the original facial features while adding realistic textures and natural lighting.

Batch: 10–20 imagesApproval Required
👔
C3

Outfit Generation

Outfits are generated independently, guided by the character's consistency profile — including preferred color palette, style notes, and forbidden items.

Batch: 10–20 imagesApproval Required
🧍
C4

Full-Body Composite

The selected photorealistic face and chosen outfit are composited into a complete full-body image. The AI handles natural body proportions and posing.

Batch: 10 imagesApproval Required
C5

Detail Refinement

The final full-body image goes through a refinement loop where you provide text instructions and the AI iterates until you're satisfied.

Batch: 5–10 imagesApproval Required
🎯

Consistency Profile

Each character has registered key traits, forbidden items, a color palette, and style notes. This profile is automatically applied at every generation stage to maintain visual coherence across the entire pipeline.

Script Pipeline

Draft, Transform, Translate

The script pipeline has three stages: draft the base script, apply a speech-style transform, then translate. At each stage the AI produces multiple versions for comparison.

S1

Base Script Drafting

Input

Plot summary, situation, character voice profiles

Output

Multiple draft versions (base language)

Approval

Select version

S2

Speech-Style Transform

Input

Selected draft + style preset

Output

Style variants (romantic / drama / cinematic)

Approval

Select style

S3

Translation

Input

Selected style version

Output

Translated script + 10-min split data

Approval

Final review

10-Min Clip Splitting

Intelligent Dialogue Segmentation

Dialogue is automatically split into 10-min clips. When three or more lines are present, these strategies are applied:

🎬

Same-Speaker Continuation

Switches to an extreme close-up (ECU) and continues filming the same speaker.

When one character's dialogue runs long

👁️

Partner Face Only

Shows the listener's face while the speaker's voice plays as an overlay.

When a reaction shot is needed

🏞️

Landscape Cut

Inserts a landscape shot as an emotional transition between dialogue beats.

When the scene mood needs to shift

[Auto Split 10min][Same Speaker Continue][Partner Face Only][Landscape Cut]
Video Pipeline

Seven Shot Types

Video production supports seven distinct shot types. Each shot can be generated independently while maintaining visual consistency through shared assets.

V1

Wide Shot (2-Person)

In: Female + male character images

Out: 10-min wide video candidates

V2

Frame Extraction

In: V1 video result

Out: Key frames (auto 12 or manual)

V3

Over-the-Shoulder (OTS)

In: Character images + wide frame

Out: OTS frame candidates

V4

Dialogue Shot

In: Speaker + OTS frame + lines

Out: 10-min dialogue video

V5

Silent Shot

In: Same as V4 (no dialogue)

Out: 10-min silent video (emotion)

V6

Continuation (ECU)

In: Previous ECU frame + images

Out: 10-min continuation video

V7

Landscape Cut

In: Character image → generate scene

Out: Clean landscape frame/video

Typical Dialogue Scene Workflow

Recommended order for a standard two-person dialogue scene

1

V1 — Wide Shot

Generate the establishing shot

2

V2 — Frame Extraction

Extract key frames as reference

3

V3 — OTS Framing

Generate over-the-shoulder compositions

4

V4 — Dialogue Shots

Produce the conversation clips

Shot Prompt Builder

Fill in structured fields — the system automatically composes optimized generation prompts.

Shot TypeWide / OTS / ECU / Landscape
CameraLocked-off / Static
ActionSit / Stand / Walk + continuity
EmotionEmotion preset + narrative text
DialogueDialogue mode / Silent mode
Pure Silent SceneLocked CameraOver-the-ShoulderExtreme Close-Up
Studio Screens(Beta - Will be Launched Soon!)

Purpose-Built Workspaces

Dedicated screens optimized for each production stage. Every screen brings together the tools and information you need for that particular task.

📊

Dashboard

Your daily production overview

See in-progress tasks, failures, daily costs, and recent projects at a glance. Your starting point for prioritizing work each day.

🏠

Project Home

Per-project hub for all resources

Manage characters, scripts, and scenes linked to a single project. Track progress and visualize asset relationships.

👤

Character Studio

Five-stage character workspace

Batch generate, review the result grid, pick favorites, and advance through each pipeline stage.

🗂️

Asset Library

Search and reuse any asset

Tag-based search, similarity search, and lineage viewer. Trace the full history of any asset.

📦

Export Center

Package finished work

Export CapCut-ready packages with organized file structures, SRT/CSV subtitles, and shot ordering.

Quality Assurance

Built-In Quality Management

The AI pre-analyzes quality to speed up the selection process. While the final call always rests with a human reviewer, automated scoring guides your decisions.

Automatic Quality Scoring

0/100

Quality Score Example

Every generated asset receives a quality score from 0 to 100. The result grid sorts highest-first to minimize selection time.

Automated Detection Flags

Bad Hands

Image

Detects abnormal hand generation: wrong finger count, unnatural joints

Bad Legs

Image

Detects leg proportion or joint placement anomalies

Face Drift

Image / Video

Flags when the generated face diverges from the original reference

Audio Music Detected

Video

Identifies unintended background music in generated video clips

Retry Policies

1

On Generation Failure

Automatically retries up to the configured limit on network or API errors.

2

On Low Quality Score

Generates new candidates to replace results below your quality threshold.

3

Manual Retry

Select individual tasks in the Render Queue and trigger re-generation anytime.

🔗

Asset Lineage Tracking: Every asset records its full creation history — prompt, seed value, AI model, and parent assets.

Use Cases

Real-World Production Scenarios

From full episodic workflows to high-volume short-form content and multilingual localization.

Step 1: Character Creation

Run the full C1 through C5 pipeline for both leads. Set up consistency profiles so characters look visually harmonious.

Step 2: Scriptwriting

Draft the base script in S1, apply a "drama" style transform in S2. Use the 10-min auto-split to verify pacing.

Step 3: Scene Design

In the Scene Builder, map out shots. Assign V1 (wide), V3 (OTS), V4 (dialogue) for conversations, and V7 (landscape) for transitions.

Step 4: Video Generation

Batch-generate videos following the shot plan. Pick the best candidate for each shot, manually retry any that need improvement.

Step 5: Export

Package finished shots as a CapCut-ready bundle — complete with SRT subtitles, CSV shot list, and organized folders.

Ready to Create?

Experience AI-Powered
Production in Action

See how characters, outfits, and backgrounds come together to create stunning AI-generated video content. Try the interactive demo now.