The Ghostface AI trend is a Halloween AI video meme in which you lie on a pink satin bed in a 90s bedroom, chat on a corded phone, and get dragged off the bed by a masked figure in a black robe and white screaming mask. To make it, upload one photo of yourself to videosi and download an 8-second silent MP4 a few minutes later.
This guide answers the questions the page FAQ does not: how the 2025 wave differs from the 2026 comeback, why the bedroom looks the way it does, what a one-step video model changes, and what to check before you post.
Key takeaways
- The original 2025 version was a still image made in Google Gemini; the video version is a newer step on top of it.
- Photo choice matters more than anything else: a clear, front-facing photo in even light, with your outfit as you want it on the bed.
- The clip is silent, so add a phone ring, a scream or a trending sound when you post.
- A standard 8-second video (Standard model, 720p) is 7 credits; the Single pack is $6.99 for 10 credits and plans start at $29.90 a month.
- Label the post as AI-generated and only use a photo of an adult who agreed.
In this guide
- How is the 2026 comeback different from the 2025 wave?
- Why a pink satin bed and a corded phone?
- Is a two-step workflow different from a one-step one?
- How do I pick the right photo?
- What is the role-reversal version?
- Is the masked figure the film character?
- What sound should I put on it?
- How do I label it as AI?
- Whose photo can I use?
- What do I do with it after Halloween?
How is the 2026 comeback different from the 2025 wave?
The first wave was about pictures. In late September 2025, people uploaded a selfie to Google Gemini, asked for a 90s bedroom with a masked figure in the doorway, and posted the stills, often as a swipe through several variations. The Tab described that version as a photo of you on a bed with Ghostface standing in the doorway behind you. The earliest video tutorials we found are dated 28 and 29 September 2025.
The October 2026 comeback is about motion. The clips we saw start on 11 October 2026 and show the whole scene playing out: the call, the rush, the grab and the drag. That is the version the Ghostface AI trend page makes. No single original creator is identified for either wave, so we do not credit one.
Why a pink satin bed and a corded phone?
Because the joke is a genre. The setting borrows the look of a 90s slasher opening: a teenager alone at night, a landline, magazines and popcorn, and a dark doorway that the camera and the audience can see but the character cannot. Pink satin makes the contrast sharper. It is the least threatening surface imaginable, so the black robe in the doorway reads instantly.
There is a practical side too. Satin creases and slides when someone is dragged across it, which gives the video model something visible to animate, and the corded phone gives your hands a job in the first three seconds. On videosi the bedroom is fixed on purpose, so every clip has the same recognisable look and you only decide who is on the bed.
Is a two-step workflow different from a one-step one?
Yes, and the difference is mostly about how many tries you pay for. The 2025 recipe used Gemini for the still and a separate video model, Kling, for the motion. That works, but each stage can fail separately: the still may change your face, and the video may let the figure just stand there or hug you instead of grabbing your ankles.
The one-step route sends your photo and a short motion reference to one video model together, so the lunge, the ankle grab and the drag are copied from the reference rather than guessed from words. videosi does this with ByteDance's Seedance. The practical results are fewer wasted renders and one place to download the file. If you enjoy the craft of prompting your own stills, the two-step route still gives more control over the image itself.
How do I pick the right photo?
Treat the photo as casting. Your face, hairstyle, build and the full outfit come from it, so choose what you want to be wearing when the figure arrives.
- Face and shoulders clearly visible, front-facing, in even light.
- No sunglasses, heavy filter or hand over the face.
- An outfit that suits a night in: a hoodie or pyjamas read better than a formal jacket.
- One person only. A group photo gives the model nothing to decide on.
Expect the drag to be the weakest second. The motion is fast and your face is smallest in that moment, so a sharp source photo is the best insurance. A failed, timed-out or blocked render returns its credits automatically.
What is the role-reversal version?
It is the 2026 variation in which you drag the masked figure off the bed instead. It works as a reply to the original, which is why it spread quickly in October. It is not offered on videosi yet; the page has one setting, the 90s pink bedroom. If you want the reversal now, the practical way is to post the original and answer it with a reaction, a duet or a text overlay in the app you post on.
Is the masked figure the film character?
No. Ghostface is the masked killer from the Scream films, and the name is how people search for the trend. The figure in the videosi clip is a generic stranger: a tall person in a loose black hooded robe, black gloves and a long white screaming mask. There is no weapon and no blood, and it is played for comedy-horror, scary but not violent.
Two consequences follow. Small details of the mask can differ from shot to shot, and you cannot put someone's face behind the mask, which is deliberate. videosi is independent and not affiliated with the Scream films. Do not describe your post as official footage or fan-film material from the franchise.
What sound should I put on it?
The clip is silent, and no single sound defines the trend. Posts use horror stings, phone rings, screams and whatever is trending that week. A simple structure works well for 8 seconds:
- A ring or a 90s-style hold tune under the first three seconds of the phone call.
- A sharp sting on the rush at about 3 seconds.
- A scream on the grab at 4 seconds, then silence for the empty room.
Add the sound inside TikTok, Instagram or CapCut, where the licensed library is, rather than baking it into the file.
How do I label it as AI?
Turn on the AI-generated label in the app you post on, and say so in the caption. The clip looks photoreal, and a label protects you from the fair complaint that you tricked someone. Do not rely on detectors either way. Google's Gemini help page says its SynthID check recognises only content made with Google AI tools, and that a missing watermark does not prove something is real. That is the reason for a plain, human label: no tool can be counted on to do it for you.
Whose photo can I use?
Your own, or an adult friend or partner who agreed to it. videosi does not accept photos of celebrities, public figures or minors, and users must be 18 or older. The prank format makes consent more important, not less: a friend who is dragged off a bed in a video they did not expect may not find it funny at a school or work account with a large audience. Ask first, and show them the result before you post it.
What do I do with it after Halloween?
Post before 31 October, when the format is at its most recognisable, and keep the file. A good, well-lit photo of yourself is worth keeping on your phone. If this clip performs, make a second one with a friend's photo as a sequel rather than waiting for the next trend. The trend itself will fade after the holiday, but a horror-comedy clip in the 90s slasher style stays recognisable well into the next year.
When you are ready, open the Ghostface AI trend page, upload your photo and press Generate. A standard 8-second video is 7 credits, and credits never expire.
