How to Do the Migos AI Video

How to do the Migos AI video is a file problem. Two stills. A template. A 9:16 download. The social posting lives on the other how-to. This one stays in the generator.

The staging rules and the original clip are on Migos AI.

How to do the Migos AI video from two photos

This is the click path on any of the public tools that published a Hotel Lobby or COLORS preset in 2026.

Step 1. Two photos, not one. Media.io’s first line is “Upload Two Front-Facing Photos.” Summrs wants face and upper body, well lit. AIReel wants one well-lit, front-facing photo per person. If you only have a group shot, stop and retake.

Step 2. One subject per file. Left person in file A. Right person in file B. Imageat says it replaces both performers. It cannot split a hug.

Step 3. Pick the room. Hotel Lobby hallway (redder, music-video). COLORS booth (flat orange, hanging mic). Summrs also splits standard vs extended. Standard is the core moment. Extended keeps more of the performance.

Step 4. Generate. Ai Miracle’s last step is a vertical 9:16 clip. Media.io says seconds. Summrs says a few minutes. Do not close the tab.

Step 5. Check the faces. Watch the whole file. Side swap, melted mouth, extra arm: throw it out. The useful check is this one. A second credit is cheaper than posting a broken clip. Loop it once with the sound off. You will catch a swapped side faster.

Step 6. Download. Media.io, AIReel, and Summrs name TikTok, Reels, and Shorts. Save locally. Then upload. A watermark means you are still on the free path.

That is the file. Fifteen seconds is enough. Imageat lists 15 or 30, and 480p or 720p. Dreamina’s sample prompt is 15 seconds. Starrd is listed as 15 seconds, 9:16. If the tool offers 30, use it only when both faces stay put for the first 15.

Prompt path, if there is no preset

Some apps have no Hotel Lobby button. Then the work looks like Dreamina’s text. Write the room. Write the mic. Write the locked camera. Do not paste a paragraph about Quavo.

Say: vertical 9:16. Performer A full body left, performer B full body right. Burnt-orange or terracotta studio. One black microphone at center. Locked, centered wide shot. Feet visible. No cuts, no zooms, no text, no logos, no extra people.

Say they take turns. A talks, B nods, then they switch. Forbid face blending and side swaps.

Do not name Quavo. Do not name Takeoff. Do not ask for their clothes. The Migos AI prompt section is the longer version of that paragraph.

Lighting that holds

Lighting that holds a face: window light or a lamp in front of the person. Not a kitchen downlight that puts caves under the eyes. Take off the sunglasses. Put the phone at chest height, not on the floor. If you only have overhead light, step toward a wall and bounce it. The model needs cheeks, not a hat brim.

Pets: one animal per frame, looking at the lens if you can bribe them. The model needs a face, even a cat face.

Clothes can be different. Height can be different. The template should keep both. If it clones one outfit onto both bodies, the photos were too similar or the model merged them.

What not to render

Copyrighted COLORS footage. A celebrity. A private photo. A screenshot of the official video with your head pasted on. The way out of that mess is the Dreamina line: permission on faces, photos, footage, music, and voices. If you cannot name who agreed, do not upload the still.

Voice models are a different file. Jammable’s Takeoff page is not this workflow.