OrnithStudio / FLUXBeauty study ↗
Render controls · 4 × RTX PRO 6000

Shape the next frame.

Set the whole render, inspect its plan, and keep a reusable configuration. Runtime connection and FLUX TP4 remain to be verified.

Prepare a render

The image

Prompt and directions stay separate in your saved plan.

Explicit dimensions below take precedence over the ratio fallback. Presets may supply prompt shaping and CLI-default steps/guidance.

Burst is a real CLI option. With a fixed seed every output repeats that seed; leave Seed empty to increment from Start seed. Sync mode waits between burst frames; async mode can queue them.

Prompt shaping · all supported controls

The runtime

Requested settings, not a report of a loaded worker.

Checkpoint, revision, precision and TP are preparation metadata; this CLI has no render flags for them. Its existing worker does not implement TP4.

Draft sets 18 steps and a 768 × 768 ratio fallback. Explicit dimensions still win. Dry-run and async are your choices. Exporting or copying a plan never executes it. Async can queue a burst without waiting; Beauty’s separate study profile keeps one frame in flight.

Advanced supported render parameters

Only actual render flags are accepted. Unknown keys, duplicate JSON keys, and overrides of active form fields are rejected.

Before a frame starts

Confirm the FLUX artifact and TP4 adapter, then measure co-resident peaks with Ornith and Gemma. Keep ≥5 GiB free on every GPU.

Dimensions use this planner’s 16-pixel grid. The current render CLI declares no verified maximum resolution; larger frames need a peak-memory check.

Beauty has two jurors: Ornith 397B and Gemma 12B. The assistant accelerates Gemma. Your champion and publication choices are final.

Your prepared render

Add a prompt to inspect the plan.

No worker is connected here. These controls prepare settings and your selected CLI command; they do not submit jobs, allocate GPUs or publish images.