Genie 4 is a world model from Google DeepMind, announced on 27 October 2026. The research post says it simulates what happens next in a scene, including how the world sounds and how people in it move and speak. You steer it while it runs. It is not a face-swap tool, and HiVideoStudio does not sell it. This page has no sample file, no credit price, and no Generate button.
HiVideoStudio is not affiliated with Google or DeepMind. Genie, Gemini, and SynthID are their names, used here only to identify the preview.
What the research preview says
The DeepMind post is by Raia Hadsell, Diego Rivas, and Hang Qi. It says Genie 4 is natively multimodal: music, ambient noise, and interactive dialogue are generated with the picture. Keyboard controls move the view. A text prompt can change the weather, move the camera, or add a person or object, and the post says Gemini adapts that prompt to the scene.
The post names Gemini Omni as the anchor for appearance and behavior, and Gemini 3.8 Flash as a guiding agent that watches frames, splits a goal into steps, and checks that each step happened. A separate experiment in the same post uses Gemini 4 Argon to break a robotics task into steps that Genie 4 renders. The example they show is a humanoid playing checkers. That is a research demo in the post, not a product on this site.
DeepMind says the preview went to trusted testers first, before a wider release that starts with Google AI Ultra subscribers in Project Genie and early-access partners of the Gemini API and Google AI Studio. TechCrunch's 27 October 2026 report, by Rebecca Bellan, says the wider rollout was described as the coming weeks, and that the model can hold a consistent scene for up to two minutes, twice the limit they attribute to Genie 3. The DeepMind post's own limitation section does not use that two-minute figure. It says visual detail and object permanence can still drift as a session runs longer, after complex edits, or while moving through the scene, and that audio accuracy, legible text, spatial accuracy, and long-term consistency are still research. Both sentences are in the sources below. They are not a benchmark this site measured.
The same post says video and audio from Genie 4 include SynthID. It also says testers in film, journalism, education, games, and engineering were trying the preview. One tester note is a single archival photograph becoming a simulation of that place. That is still a simulation of an environment. It is not a tool that pastes one person's face onto another person's performance.
Echo-3 is the other October world model
Echo-3, in the posts checked for that page, takes a phone video of a real room and builds a 3D Gaussian splat. PlayCanvas rendered one of those splats. Echo-3 does not steer a generated world with keyboard and prompt the way the Genie 4 post describes, and the Echo-3 posts do not describe native dialogue. Genie 4 does not claim to measure your room. The long version is Echo-3 and Genie 4.
What a creator can make here
A HiVideoStudio clip is a photo in a template, not a world you walk through.
- Survival Horror AI: one photo, 15 seconds, 50 credits, rainy flashlight scene.
- AI Zombie: one photo, 15 seconds, 45 credits, three templates.
- Ozil Goal AI: one photo replaces the scorer in a fixed move, 50 credits for 15 seconds.
- Thriller Dance: one photo, 15 seconds, 50 credits, silent.
Henge can retarget a guide video you already have. It is not Genie 4, and it is not for sale here either. Face-swap and character-swap searches are sorted on AI character swap video.
How to tell Genie 4 from a face swap
- Read the input. Genie 4's post describes prompts, keyboard controls, and a simulation. A face swap starts from a person's photo and someone else's footage.
- Read the output. Genie 4's post describes a world that keeps going and makes its own audio. A face swap keeps the original performance and changes the identity.
- If you want a 15-second clip from one photo, use an effect this catalog sells. Genie 4 is not on that list.
- Do not send a portrait to this page. Nothing here is generated.
Quick answers
Is Genie 4 a face-swap tool? No. The 27 October 2026 research post describes a steerable simulation with native audio.
Can I use Genie 4 on HiVideoStudio? No. This page is not for sale. DeepMind's post says the preview started with trusted testers.
How long does a Genie 4 scene stay consistent? TechCrunch reported up to two minutes, and said that was twice Genie 3. DeepMind's post says detail and object permanence can still drift on longer runs. This site did not measure a session.
What is Echo-3, next to this? A reconstruction of a captured place, rendered in one demo by PlayCanvas. It is not this preview.
Sources
Google DeepMind, "Genie 4: A new generation of world simulation," 27 October 2026. TechCrunch, Rebecca Bellan, 27 October 2026. Google's announcement post, 27 October 2026. Last checked 2026-10-29.