Higgsfield inside Blender: you move the camera yourself, with your phone

For years we left the camera to chance when generating video with AI: you write the prompt, cross your fingers and burn credits until something resembles what you had in your head. The Higgsfield Blender Plugin puts a piece of software in between — Blender — and with it you get back what matters to a cinematographer: the rail, the lens, the blocking and the timing are your call, and you operate the camera with your phone.

Higgsfield’s dialog below the viewport: scenes, models, character animation, image, video and camera.

The facts

• The plugin is free and installs from Higgsfield’s plugins section. Drag it into Blender or ask your agent: I just told Claude “install this plugin” and that was it. Codex works too.
• Inside Blender you get a panel with scenes, 3D models, character animation, image, video and camera.
• Text-to-3D with Meshy: “create a Formula 1 car” and the model appears in the scene. For complex environments I used GPT-6 Astra, the most advanced model available today: “create a racetrack for NASCAR-type vehicles” and it modeled the whole environment.
• Image from the viewport: you generate from what you see in Blender, with Blender’s own camera angle, lens and movement. As specific as you want.
• Camera with your phone: connect your iPhone with a QR code, do the move by hand, it records, and that position and motion data drives the video generation in Seedance 2.5.

What I liked most: the camera

I did the move badly on purpose — I’m learning — and it still works: the video comes out with a live, hand-operated feel, like a helicopter following the car, and that gives it a realism a prompt won’t. I didn’t work the texture or the material of the vehicle; I was only looking at camera position and movement. That’s the point: not generating everything from scratch, but putting Blender in the middle to get specific controls, scene design and movement design — and bringing textures and tools from Higgsfield straight into Blender.

The move I recorded with the phone, applied to the model generated with Meshy.

From previz to shot

My practical use is previsualization with directors. You model the track, place the cars, pick camera positions and generate: I did one Formula 1 and one NASCAR without texturing anything. That’s not the final finish — it’s the previz of your environment, your models, the pieces you’re going to play with before building the scene. From there you take it to Seedance or Genjutsu and keep going.

And it’s not all blockbusters. Someone studying biotechnology wrote to me saying they want to do science communication: with this they illustrate their videos. I think that’s one of the best uses.

The environment I asked GPT-6 Astra for, ready to texture and place cameras.

The LATAM read

I don’t know how to model in 3D. There are people who do it better, by hand and in detail, and those people are still needed for the finish. But for anyone producing here — where previz for a car sequence was a big-studio luxury — being able to block a scene in an afternoon, show it to the director and decide the rail before renting anything is real money. And the phone-operated camera is the part that’s ours: the handheld gesture belongs to us, not to the model.

My personal opinion — Daniel Blanco A.M.C.

Much of what gets said about generative AI is that there’s no specific control. I think there is, and there are many ways to get there; this is one of the most interesting. Everything is getting simpler every day and giving us much more specific controls. And what’s coming excites me more: there are already models that accept ProRes and 16-bit, very specific things that are going to integrate into cinema — I’ll make a video on that soon.

Where to start

Install the plugin, ask Meshy for a simple model and connect the phone. Do a single camera move — one you’d do on a set — and generate. If the result feels operated by a human, you already understand what it’s for. This video falls short of everything that’s in there; I’ll make specific videos on Blender and on how GPT-6 Astra enters Higgsfield for models, games and apps.

Images: frames from Daniel Blanco A.M.C.’s video.

Original source: Blender + GPT-6 Astra + Seedance: Higgsfield’s free plugin, on YouTube

Extra for you: if you’re experimenting with AI for video, my Higgsfield link has discounts and promos to get started — and using it supports this content.