SIMBA OMNI 🐶 // OUR OWN RENDERER

Simba Omni — our pipeline, our agents.

Most tools hand your prompt to somebody else's video model and hope. Simba Omni renders through weights we trained, and the brief it renders is written by a chain of agents that has actually seen your product.

9:16 vertical 5–15 seconds 768P source Trained adapters

No credit card required · Free sandbox included · Every clip comes out reel-ready

Live Interactive Demo

Watch Simba Omni and the orchestration preset run

Two ways into the same pipeline: animate a finished still into a vertical ad, or start further back and let the agents plan the photos and write the words first.

simba.orchestration // live-sandbox
MONITORING
ACTIVE STAGE

ORCHESTRATION LOG STREAM tnum // 480ms step-cycle
Initializing Simba orchestration preset...
The Model

One renderer, reached from both ends

Simba is a family, not a single button: one renderer with different trained adapters, offered wherever a video can start from.

References → video

Simba Omni

The renderer without a trained look, so it follows your brief instead of a style. Reach for it when the product is not clothing.

References → video

Simba Fashion

Our own renderer, trained on how a presenter moves and shows a garment. The one to use for clothing.

Still → video

Simba Omni · Animate

Tuned for real skin and hair at phone size. The steadiest of the three, and vertical only.

What every Simba render is

Shape
Vertical 9:16, because every advert this product exists to make is a vertical one.
Length
5 to 15 seconds per clip — the range a reel is cut from, priced per second.
Resolution
768P at the source. The composer scales every frame to 1080×1920 regardless, so higher rungs buy pixels the finished file throws away.
Weights
Adapters we trained and pin by hash, not a general-purpose model with your prompt bolted on.
The Orchestration Preset

Nineteen agents, one brief, in order

The preset is the part nobody sees and everybody feels: each agent is asked one narrow question, hands its answer to the next, and returns a schema rather than a hope. Every stage runs on its own queue.

Stage 01

Read the product

The photos are attachments, not filenames. One dense technical description comes out, and everything downstream is written against it.

ProductAnalysisAgent ProductClassifierAgent
Stage 02

Plan the frames

A set is worth more than one photo repeated, so the plan says what each frame is for before a single pixel is rendered.

ShotPlannerAgent SceneScoutAgent SceneReviewAgent
Stage 03

Direct the video

A storyboard is a sequence, so every clip is written to hand over to the next — then the motion, the script and the delivery are decided per clip.

VideoStoryboardAgent MotionDirectorAgent VoiceScriptAgent VoiceDeliveryAgent
Stage 04

Write the words

Three ad angles, the whole landing page, and the script the call centre reads when the phone is answered.

AdStrategyAgent AdCopyWriterAgent LandingPageCopyAgent ConfirmationScriptAgent
Transparency

What "Simba Omni" refers to

Simba Omni is the name of our orchestration engine — the agent workflows, the prompt construction and the adapter pipeline that turn a product brief into finished media. It is not an AI model in itself. Under the hood it coordinates several: adapters we train ourselves, open-weight models, and hosted commercial ones, each picked for the job it is best at.

Every Simba name works this way. The name is ours and stays stable for you; the models behind it are chosen on merit and change as better ones arrive.

Orchestration, not a model Multi-model by design Our adapters + open weights Chosen per job

We use cookies

Essential cookies only — to keep you logged in and secure your session. No tracking or advertising cookies.