From sets to camera moves, Higgsfield integrates with Blender

The Higgsfield add-on lets users build sets, animate cameras, and generate assets inside Blender while keeping the resulting scene fully editable.

Higgsfield is continuing its push into professional creative software with a new add-on for Blender. It brings scene, 3D model, animation, image, and video generation into the software’s interface. The tools can be controlled from a floating bar in the 3D viewport or through an MCP-connected agent.

The most significant feature focuses on scene preparation. Users describe a set, its objects, their placement, and the desired lighting. Scene Builder then creates a blockout directly inside the open Blender file. Unlike a simple reference image, the result consists of actual scene elements: the geometry can be moved, resized, deleted, or manually replaced.

This initial construction is intended as a working foundation rather than a finished result. It can be used to test framing, character movement, or the overall balance of a set before spending more time on detailed modeling. A new instruction can also reorganize the entire shot within seconds, without preventing users from refining each element afterward with Blender’s standard tools.

The add-on includes seven sections: Scene Builder, 3D Model, Character Animation, Image, Video, Camera, and Asset. The 3D module accepts either a prompt or a reference image and places the generated model at the 3D cursor. Higgsfield says the output includes materials and textures, with Meshy 5 currently listed as its 3D generation model.

Character Animation adds a rigged character and places keyframes on the timeline based on a natural-language description of the movement. The result uses Blender’s standard bones and actions, allowing users to adjust the animation, edit the curves, or change the timing as they would with conventional motion-capture data.

Camera movements can also be directed through natural-language instructions. Users can describe a tracking shot, an orbital movement, or a handheld approach and receive an editable animation. Another feature uses a phone’s physical movement to control the scene camera in real time. The approach is primarily designed for rapid shot previsualization rather than replacing traditional cinematography work.

Higgsfield then connects the 3D viewport to its image and video models. A scene can provide the starting composition for a finished image or be sent to Seedance 2.5 to generate a video sequence. Results can return to Blender as an object, plane, texture, or file accessible through the integrated asset library.

The integration relies on two connection methods that should be distinguished. Higgsfield’s standard MCP server allows a compatible assistant to generate assets. The new Blender Bridge additionally gives the agent access to the installed add-on and the currently open scene. An agent can therefore position an object at the 3D cursor, build a blockout, or launch a video generation without working on an isolated file.

The connection uses `https://bridge.higgsfield.ai/mcp`, which must be added as a custom connector in a compatible client. Higgsfield lists Claude, Claude Code, OpenClaw, Hermes Agent, and NemoClaw among the tools supported by its MCP infrastructure. Its Supercomputer platform can also orchestrate these operations as part of a larger workflow.

The add-on supports Blender 4.2 through 5.1 on macOS and Windows. It is installed by dropping a downloaded archive into an open Blender window. Generative processing takes place on Higgsfield’s servers, so a powerful GPU is not required for that portion of the workflow. Local Cycles or EEVEE rendering, however, still depends on the user’s hardware.

An internet connection and an active Higgsfield account are required to launch new generations. Elements already added to a project or saved to disk remain available offline. The credit cost appears on the Generate button before each operation and varies according to the selected model, settings, and number of variants. The add-on uses the same account and credit balance as the web platform.

The integration primarily reduces the separation between 3D previsualization and image or video generation. It does not remove the need to inspect generated models, rigs, camera movements, or materials. Instead, it adds a faster preparation layer whose results remain open enough to be manually corrected and developed further.