Scope Flux RT

Scope plugin for TensorForger FluxRT, adapted as native Scope video pipelines.

This package vendors the FluxRT model internals needed for in-process Scope execution and intentionally excludes FluxRT's demo runtime, multiprocessing StreamProcessor, shared-memory transport, and output scheduler. Scope owns frame scheduling, graph routing, lifecycle, and playback.

Pipelines

• flux-rt: FLUX.2-klein-4B with FluxRT spatial cache.
• flux-rt-int8: same pipeline using the FluxRT int8 transformer/text-encoder artifact.

Model Artifacts

Scope downloads artifacts into DAYDREAM_SCOPE_MODELS_DIR or ~/.daydream-scope/models:

• black-forest-labs/FLUX.2-klein-4B
• aydin99/FLUX.2-klein-4B-int8 for flux-rt-int8

The FLUX.2 model may require Hugging Face authentication and acceptance of upstream terms.

Install

Desktop App

1. Open Settings → Plugins tab.

2. Paste the following into the installation input field:

   https://github.com/daydreamlive/scope-flux-rt
   

3. Click Install and wait for the server to restart.

The plugin and its flux-rt / flux-rt-int8 pipelines will appear automatically once the restart completes. See the Plugins guide for more detail.

Manual Installation (CLI)

Follow the manual installation instructions for Scope, then install the plugin from the scope directory:

uv run daydream-scope install https://github.com/daydreamlive/scope-flux-rt

Confirm the plugin is installed:

uv run daydream-scope plugins

Confirm the flux-rt and flux-rt-int8 pipelines are available:

uv run daydream-scope pipelines

Upgrade

Desktop App

Open Settings → Plugins tab. If an update is available, an Update button appears next to the plugin — click it and wait for the server to restart.

CLI

uv run daydream-scope install --upgrade https://github.com/daydreamlive/scope-flux-rt

Uninstall

Desktop App

Open Settings → Plugins tab and click the trash icon next to the plugin.

CLI

uv run daydream-scope uninstall scope-flux-rt

Usage

Use Scope's native prompt control or the graph prompts input to drive FluxRT. The plugin does not expose a separate default prompt field, so there should only be one prompt input in the UI.

Set reference_image_path at load time if you want FLUX.2 reference-image conditioning. Load flux-rt for the base model or flux-rt-int8 when you have the int8 artifact available.

Licensing

FluxRT is distributed under the Unlicense. Some vendored files retain Apache-2.0 headers from Black Forest Labs, Hugging Face, and InstantX-derived code; those notices are preserved in the source files.