Can ChatGPT, Gemini or Sora Do the AI Zombie Trend?

The trend is a short vertical video. One person has turned into a zombie, and the other recognizes them. Mid-air, the clip cuts to a warm memory of the same two people. The earliest large post we found is an Instagram Reel from 26 September 2026 that credits the original idea to @techmaf. A Russian tutorial on Habr from 30 September describes the same shape as three scenes (threat, choice and warmth). Its template offers 9:16 with 5, 20 or 30 seconds.

So a tool has to do four things. It has to keep two real faces recognizable across both parts. It has to give you a vertical clip. It needs sound, because the piano note on the cut matters. And it has to run long enough for the run, the leap and the memory: our own story takes 15 seconds. The full trend guide explains the four beats.

Not the video. OpenAI's help article says ChatGPT Images lets you create new images and edit existing images in ChatGPT. You can upload an existing image and describe the changes you want. It lists ChatGPT Images as available on all tiers. The article does not mention video, and we found no OpenAI page that announces video generation inside ChatGPT as of 15 October 2026.

ChatGPT is still useful for the trend, in two ways. It can write the prompt for a video tool. It can also make the still first frame of two people in the cold scene, which you then hand to a video tool. The trend guide shows where that still goes. But the moving, 15-second clip has to come from another tool.

No, because Sora is closed. OpenAI's help center says the Sora web and app experiences were discontinued on April 26, 2026, and the Sora API will be discontinued on September 24, 2026. The company announced the decision on 24 March 2026, when Engadget reported that OpenAI had "decided to discontinue Sora in the consumer app and API".

If a guide tells you to start the zombie hug in Sora, check its date: the app it describes no longer exists.

Yes, in short clips. Google's Gemini Omni page says you can create 10 second videos with native audio generation and turn up to 5 photos into a video. Google's launch post of 19 May 2026 says Omni Flash is rolling out to all Google AI Plus, Pro and Ultra subscribers globally. You need a paid Google AI plan, and the page says you must be 18 or older. Google says every video made in the Gemini app carries an invisible SynthID watermark.

Two photos fit in the 5-photo limit, so the two-photo setup of the trend works on paper. Three limits matter for this trend:

• Length. A new generation is 10 seconds at most. Google's August update lets you extend a video in 10-second increments up to a total of 40 seconds, so you build the zombie hug in steps.
• Shape. The help page says you choose the aspect ratio before the video is created. The default for text prompts is landscape, and an uploaded image or video sets the ratio to match it, so pick 9:16 or upload portrait photos.
• Flow. In Google Flow, ingredients-to-video with Veo 3.1 Lite or Fast is limited to 8-second videos, while Gemini Omni Flash 1.1 offers 4, 6, 8 or 10 seconds. Either way, you join the clips yourself.

We have not tested how well Gemini keeps two faces across the cold part and the memory, so we make no claim about it. Try a short clip with your own photos before you plan a whole video.

Zombie Hug uses Seedance by ByteDance. ByteDance's launch post of 12 February 2026 says users can input up to 9 images plus natural language instructions and that the model supports 15-second high-quality multi-shot audio-video output with dual-channel audio. That is why our video is 15 seconds long. The AI models page has the details of our setup.

Every tool has rules for photos of real people, and the zombie hug uses two. The rules we found in official pages:

• Gemini. Google's help page says only upload photos you have the right to use, and that the feature is not available to users under 18.
• Seedance. ByteDance says that to use real human portraits as subject references, identity verification or prior legal authorization is required.
• Zombie Hug. Both people must be 18 or older and both must agree before you pay. We do not accept photos of children or famous people.

When you post the result, label it. TikTok's synthetic media policy says creators must label AI-generated content that includes realistic images, audio or video. Meta has renamed its "Made with AI" label to "AI info" across its apps. It adds the label when people disclose that they are uploading AI-generated content.

Use Gemini or Google Flow when you want to direct the video yourself: your own story, your own prompts, your own length. Use Zombie Hug when you want the finished zombie hug from two photos without prompts or editing.

What we measured ourselves:

• Spec. We checked our 7 sample clips with a media tool on 15 October 2026. Each is 720x1280 (vertical 9:16), 15.04 seconds long, with a sound track. See them on the examples page.
• Speed. On 15 October 2026 we timed two videos. They took 5 minutes 11 seconds and 4 minutes 39 seconds to make. Most videos are ready in 5 to 10 minutes.
• Price. One video is $6.90. A failed video costs nothing, and every video includes 1 free remake within 7 days.

The cut in mid-air and the warm memory are built in, and you choose who turns and one of 3 scenes (ruined street, pumpkin patch or marigold field). You cannot write a prompt, and that is the trade-off.