PROMPTING IS DEAD.
START DIRECTING.

Orchestrate individual scenes with surgical precision. Banana Split is the industrial-grade studio for professionals who refuse to settle for monolithic generation.

Compiling Visual Reality...

CUSTOM IDENTITY STUDIO

Lock down visual invariants permanently. Our forensic engine analyzes your style and compiles it into an invariant token. Type @cinematic-glow inside the Director chat to invoke.

Active Identity

Cinematic Glow

@cinematic-glow
Cinematic Glow
forensic core: activestyle locked

A dark, dramatic cinematic lighting setup utilizing a strong key light positioned 45 degrees off-camera, casting a volumetric shaft of warm amber illumination that wraps around the side of the subject. A subtle, cool teal rim light outlines the opposite shoulder, creating a sharp separation from the pitch-black, low-key background. The shapes of the light are organic and defined, featuring soft-edged shadows that contour the facial structure. The subject is placed slightly off-center according to the rule of thirds, looking past the lens. The camera is positioned at a low angle, slowly tilting upward with a narrow depth of field, rendering the foreground elements into a soft bokeh while keeping the subject's expression in sharp focus.

skin
ResolvedActive

skin

@skin

This identity represents skin, seeking to lock in realistic, high-quality skin textures, micro-pores, and natural dermal structures across generations.

selfie
ResolvedActive

selfie

@selfie

This identity represents a selfie posture, seeking to replicate casual smartphone-held angles, candid expressions, and natural portrait focal depth.

product
Primary Identity Node: ResolvedGlobal Profile

product

@product

This identity represents a product, seeking to recreate exact, unmatched product comprehension and packaging shapes to build unified brand campaigns.

Product
01 / Product
Character
02 / Character
Combined composite render
05 / Composite Render
Clothing
03 / Apparel
Environment
04 / Environment

SEQUENTIAL TASK PIPELINES

Now you can orchestrate complex, multi-step creative pipelines sequentially and synthesize complete compositions—all in a single prompt. Just list the assets you need, and the studio will generate each one before automatically compiling them into your final scene.

Key Action Example:

"First, generate a studio shot of my product bottle. Then, create a character portrait. Next, design a custom leather jacket apparel asset. After that, create a foggy pine forest environment. Finally, combine all four assets together, dressing the character in the jacket and placing them with the product in the forest for a seamless composite render."

  1. 01.Product Asset Node
  2. 02.Character Asset Node
  3. 03.Apparel Asset Node
  4. 04.Environment Asset Node
  5. 05.Combined Composite Render

MOTION CONTROL STUDIO

Extract camera choreography and human kinetics from any source video and inject them directly into your scenes. Achieve perfect visual pacing and structural alignment with zero manual framing.

extracting kinetics01 / Source Motion
rendering target02 / Synthesized Control

CREATIVE POSSIBILITIES

manifest limitless creative variations. merge custom styles, character profiles, camera movements, and lighting structures. Banana Split provides absolute visual sovereignty.

Creative Possibility 1
cinematic shots
Creative Possibility 2
ad campaigns
Creative Possibility 3
graphic design
Creative Possibility 4
and more...

AUTOMATED STUDIO DIRECTOR

Step into the director's chair. The Director is a reasoning agent that reads your request and makes smart technical decisions to optimize and create the best possible workflow to generate the highest-fidelity results. No prompt engineering or parameter tuning required.

  • 01 /
    Video Analysis

    Feed references directly. The Director transcribes audio, dissects pacing, and extracts visual motion vectors so the engine can learn custom video structures.

  • 02 /
    Image Analysis

    Analyze warehouse uploads or results. The Director runs visual forensic breakdowns to extract stylistic, lighting, and composition tokens.

  • 03 /
    Identity Integration

    Invoke locked aesthetics, personas, characters, or products instantly using clean `@name` handles directly in your chat request.

  • 04 /
    Batch Tasks

    Send multiple unrelated prompts concurrently. The Director dispatches them together, letting you return to ready results without tab switching.

  • 05 /
    Sequential Tasks

    Chain tasks together so outputs flow into subsequent steps (e.g. Image → Upscale → Video) without manual intervention.

  • 06 /
    Unified Control

    Orchestrate all dimensions of production—from image and video generation to custom audio synthesis and motion kinetics transfer.

Orchestrator:Deterministic DAG Routing
State Management:Zustand Sync / Supabase
Ask the Director...
Usage:0

STATE OF THE
ART MODELS.

Banana Split connects directly to ComfyUI pipelines. We pass-through GPU execution times with zero hidden fees.

VE-3.1 PRO

Veo 3.1

High-fidelity Google Veo video generation. Perfect for atmospheric depth, complex camera moves, and organic simulation.

Max Duration:12s
Resolution:1080p Cinema
KL-3.0 ULTRA

Long 3.0 (Kling)

Ultra detailed 4K cinematic video clips, motion steerability camera pans, and advanced tracking controllers.

Max Duration:10s
Resolution:4K UHD
SD-2.0 CORE

Seedance 2.0

Consistent character movement, stable visual rendering, and complex human action flow synthesis.

Max Duration:12s
Best For:Character Motion
NB-V2 CORE

Nano Banana 2

Fast-pass draft rendering and subject adjustments. Perfect for testing layout grids and composition drafts.

Max Resolution:1024x1024
Ideal Use:Testing / Drafts
GPT-2 IMG

GPT 2 Image

Custom high-dimensional visual model optimized for direct layout processing and stylistic variations.

Max Resolution:2048x2048
Ideal Use:Creative Render
SO-2 ACTOR

Sora 2

Complex physics simulation, character consistency, and photo-real environments with integrated audio sync.

Max Duration:12s
Audio Sync:Supported
SD-5.0 EDIT

Seedream 5.0 Pro Edit

Edits and generates images from single-image or multi-reference inputs, supporting up to 10 references, aspect ratio selection, and 1K / 2K resolution tiers.

Max References:10
Resolution Tiers:1K / 2K
SD-5.0 PRO

Seedream 5.0 Pro

Generates high-quality images from text prompts, with custom aspect ratio selection, strong prompt following, and 1K / 2K output resolution tiers.

Category:Text-to-Image
Resolution Tiers:1K / 2K