Echo-3, in the posts this page cites, rebuilds a real place. It does not take a text prompt and invent a video, and it does not swap a face. HiVideoStudio does not sell it. There is no sample file on this page, no credit price, and no Generate button.
What the 27 October posts show
On 27 October 2026 at 17:32 UTC, @playcanvas posted: "Generated by the insanely powerful Echo-3 world model. Rendered by the open source PlayCanvas Engine." When this page was checked on 2026-10-29, that post had 121 likes. It quotes a post by David Novotny from earlier the same day.
Novotny's post says a quick phone video of a room was built into a 3DGS model, and the house was restyled. 3DGS means a 3D Gaussian splat. When checked on 2026-10-29, that post had 2,439 likes. It links SpAItial's announcement from the same afternoon.
SpAItial's blog the same day, "Introducing Echo-3 and the SpAItial app," calls Echo-3 a physically grounded world model. The post says a capture of a few minutes on an iPhone becomes a 3D environment with metric scale, so dimensions match the real space, and that missing surfaces are filled in to match the surroundings. It names an iOS app, a first month on the subscription, an enterprise API, and Android as not ready yet. The blog post does not link a weight file, a training repository, or an open-source license for the model.
The words "open source" in the PlayCanvas post belong to the PlayCanvas Engine, which rendered the clip. The engine repository is public and states an MIT license, and its feature list includes loading and rendering 3D Gaussian splats. That is the renderer. It is not a release of Echo-3 weights.
A watchlist note called Echo-3 an open-source omni world model comparable to Genie 4. The posts above do not show that. They show a capture of a real place, a splat, a restyle of that splat, and a web render. Genie 4 is the other tool: a simulated world you steer while it runs. The side-by-side is Echo-3 and Genie 4.
Echo-WM, the August 2026 paper arXiv:2608.23189, is a different name and a different model. It is not the Echo-3 in these October posts.
What this site can generate
A photo-to-video effect here does not reconstruct your room and does not hand you a splat. Survival Horror AI is one photo and a 15-second rainy scene at 50 credits. AI Zombie is one photo and a 15-second template at 45 credits. Ozil Goal AI replaces the scorer in one fixed move at 50 credits for 15 seconds. None of those is Echo-3.
Henge is also not Echo-3. Henge swaps or retargets a person inside a guide video. It does not build a metric model of a room.
How to read an Echo-3 clip
- Check whether the clip is a view of a captured room, or a scene nobody filmed. The posts cited here are the first kind.
- If the file is a splat in a browser, the PlayCanvas post is claiming the engine, not a video generator you prompt.
- If you wanted a 15-second clip from one photo, use an effect in the catalog. Echo-3 is not on that list.
- Do not upload a face to this page. It does not generate.
Quick answers
Is Echo-3 a face-swap or character-swap tool? No. The posts cited here describe a 3D reconstruction of a real space.
Is Echo-3 open source? The PlayCanvas Engine that rendered the demo is open source. The Echo-3 posts checked on 2026-10-29 do not publish model weights.
Can I make an Echo-3 world on HiVideoStudio? No. This page is not for sale. It has no Generate button.
How is this different from Genie 4? Echo-3 in these sources rebuilds a place you captured. Genie 4 simulates a place and lets a tester steer it, with generated audio.
Sources
PlayCanvas, 27 October 2026. David Novotny, 27 October 2026. SpAItial, Echo-3 release. PlayCanvas Engine. Like counts are from the posts when checked 2026-10-29. Last checked 2026-10-29.