The client suddenly added a render
At 9 p.m. last Thursday, I was in a Hangzhou café when a client messaged: “Can you also send a version with an effect close to the real scene?” My first instinct was to find new references, render again, and tweak the materials. My friend Lin Zhe was editing product showcase images too. He said, “Check out Blender 5.3 first; the new version supports Gaussian splatting.” I still had no idea what the feature actually did and nearly spent the whole night repeating renders.
What it is, and who is already using it
Simply put, Gaussian splatting turns real footage into a 3D scene you can view from different angles. Similar presentations used to require professional scanning equipment and complicated post-production. Now, I can import footage into Blender, then adjust the camera and framing.
Independent designer Chen Yu used it from her Chengdu studio to create an “effect preview on a kitchen counter” for a coffee machine brand. She started with real-scene photos, then moved the camera to generate different angles. The client could see the sense of space without waiting for the final video.
Still, it is not “one click turns photos into a movie.” The shooting angles need enough coverage, lighting changes should not be dramatic, and people walking through the frame may produce distorted edges. I got this wrong too: on my first attempt, I recorded only two front-facing clips. When I tried rotating toward the side, the image looked stretched thin, so I had to return to the location and shoot more. Now I record a short test first and decide whether it is worth continuing.
What replicating it costs today
I can keep the replication cost at 0 yuan and 2–4 hours. The technical bar is “comfortable with Blender’s basic controls and able to shoot a short video or photo sequence.” The first step is to open the Blender 5.3 download page on the official website, click the button for my operating system, and install it.
I would not rush to learn something just because it has “splatting” in the name. First, I would check whether the footage fits this workflow. A phone is enough for capture, but stable movement, fixed exposure, and a tidy scene matter more than an expensive camera.
If I only need a poster or a static product page, this tool may not be necessary for me. If the client already has a 3D model or high-quality renders, adding it to the workflow may not be worthwhile either. There is no harm in waiting. I would clarify the delivery scope first; that is usually safer than chasing a new feature.
Trying it at my current stage
If I am just starting out: I would first make a 10-minute sample using my own portfolio. I would not take a client order for it, because I do not want a new effect to disrupt my pricing.
If I already have 1–2 clients: I would first ask whether they need a presentation that can be viewed from different angles. Then I would test it with 20 minutes of footage and place the splat result beside a conventional render in the same proposal.
If I am scaling up: I would document the capture requirements as a handoff checklist, including the route, frame rate, lighting, and file naming. Once someone on the team can produce usable footage consistently, I would consider making it part of the regular workflow. I might also invite myself to review the first test with the team and check whether it really saves communication time instead of adding another step.