How to Make a Hug in Heaven AI Video From Two Photos

The heaven hug trend turns two still photographs into one short reunion clip: you run toward someone you miss through light or cloud, and the video ends in an embrace. You need two photos, a browser, and about ten minutes. No prompt writing, no editing timeline, nothing to install. This guide covers the three steps, the four scene variations, the photo mistakes that break faces, and the questions worth answering before you post anything.

The trend spread through TikTok and CapCut templates under names like heaven reunion, stairway to heaven, and abrazo en el cielo — the hug in heaven. TikTok's #heavenreunion hashtag alone had 2,771 videos by October 15, 2026, according to the count reported on MakeViral's template tracker, and searches spike around Día de Muertos, when families across Mexico and Latin communities make reunion clips for the ofrenda.

What exactly is a hug in heaven video?

It is an AI-generated clip built from two photographs. One photo shows you; the other shows a person — or a pet — you miss. The effect casts both faces into a pre-built scene: a stairway of warm light, a slow walk across clouds, or a quiet embrace in soft haze. Camera movement and music are built in. You choose the scene; the faces come from your photos.

It is not a CapCut template. Templates drop your photos into somebody else's edit, watermark included. A generated clip is your own take, rendered fresh from your two pictures.

What do you need before you start?

Two separate photographs, one person in each. Front-facing or slightly turned, with even light on the face and nothing covering it. Waist-up framing gives the model enough body to work with.

Old photos work — in fact they are the most common case. A faded print, a scan from decades ago, a black-and-white original: lay the print flat, photograph it with no glare, and the effect keeps the face at the age shown in that picture while rebuilding colour naturally.

If you have tried a similar photo-to-video effect before, the upload habits carry over: sharp, well-lit, single-person photos in, better takes out.

How do you make one, step by step?

Step 1 — Add your photo, then theirs. The order matters: the first photo is the person who runs, the second is the one who waits. Crop group shots down to one person first — group photographs are the usual reason a take comes back strange.

Step 2 — Pick the scene and keep 9:16. You choose between the stairway run, the cloud run, or the quiet embrace. Length and shape are set before you generate: clips run around ten seconds and export vertically at 9:16, the shape TikTok, Instagram Reels and YouTube Shorts expect, with soft music and no watermark.

Step 3 — Watch it alone, then decide. Rendering takes a few minutes. Every take is a fresh generation, so if a face drifted or the hug feels stiff, run it again with a tighter crop or better light rather than re-uploading the same photo. Download the MP4 only after you have watched it once by yourself.

The whole workflow runs in the browser and needs no prompt — the scene is pre-directed, which is why a first take usually looks like the trend instead of like a random sky. Textideo's Hug In Heaven generator runs this exact three-step flow, and renders stay private in your own library until you choose to download or share them.

Which scene should you pick?

Scene	What it looks like	Choose it when
Stairway run	You climb a stairway of warm light; they wait at the top	The reunion should feel like arriving after a long journey
Cloud run	A slow walk across golden cloud at sunset — the abrazo en el cielo version	It is for Día de Muertos, or you want it calmer than a run
Quiet embrace	No running; two people come together in soft light and hold on	The run feels like too much to watch — anniversaries, first years
Pet take	A dog or cat comes to you and the camera drops to a kneel	The second photo is a pet; two people in one photo also work

Why does my video not look right?

Problem	Usual cause	Fix
The face stops looking like them	Source photo is blurry, shadowed across the eyes, or heavily filtered	Rescan or re-photograph the print flat, glare-free; regenerate
The two faces swap or blend	One photo contains two people	Crop to one person per photo and run again
They look the wrong age	The clip keeps the age in the photograph you chose	Pick a photo from the era you want to remember
Wings, halos or stray text appear	Random take variation	Regenerate; each run is a new attempt
The run gets refused	Some photos fail screening	Try a different photo of the same person

What this effect still cannot do

Honest limits, so you know before you start:

• No custom scenes or prompts. You pick a variation, you do not direct it. If you need a specific setting — their old kitchen, a particular song — this is not the tool for that.
• Around ten seconds, one embrace. There is no story arc, no dialogue, no multi-shot sequence.
• Likeness depends on the source. A thirty-year-old sideways print will come back softer than a sharp portrait. The effect keeps what the photo gives it.
• It is not footage. Every take is AI-generated. It should never be presented as real video of the person.

Is it OK to post a video like this?

Go slowly, and decide with your family before you post. These clips are genuinely comforting for some people and unsettling for others, and relatives do not always feel the same way you do about them.

Three rules worth following: only use photos you have the right to use — never a stranger's or a public figure's without permission. Check with the people who knew them before anything goes public; many families keep the clip in the group chat rather than on a feed. And label the post as AI-generated, so nobody mistakes it for footage that was really taken.

Last updated

2026-11-02. Scene names, photo requirements, clip length and privacy behaviour checked against the effect page on that date. If the effect changes, this guide gets corrected rather than left standing wrong — if your take came out differently, especially with an old print, say so.