How to Use Sora 2 Cameos: Put Yourself in AI Videos (Full Tutorial)
Sora 2 just shipped a game-changing feature called Cameos — it lets you put yourself into any AI-generated video. Walking a bear in Tokyo. Surfing a wave. Stepping out of a billboard. With a clean Cameo capture and a sharp prompt, the lip-sync is so good people stop scrolling.
In this full tutorial I walk through exactly how to create your own Cameo, the settings I use to get the best results, what to do when generations fail, and the prompt structure for viral hooks. There's a free pack of 50 ready-to-use Cameo hooks linked above.
What are Sora 2 Cameos?
A Cameo is a 5-second video of yourself that Sora 2 uses as a reference to put you into any generated scene. The model learns your face, your voice, and your motion — then re-renders you doing things you've never actually done. It's the difference between "AI made a video" and "AI made a video of you".
Examples I show in the tutorial: me walking a giant bear through Times Square, surfing a 30-foot wave, presenting a podcast set I've never been on. All from one 5-second clip.
How to create your Cameo (step-by-step on iPhone)
You record the Cameo inside the Sora iOS app. The setup matters more than the prompt — get the capture right and 90% of your output is good. Here's what I do:
- Open the Sora app, tap + Create cameo at the bottom of the Cameos rail.
- Hold the phone at eye level, neutral background, even lighting (no harsh sun, no shadow on half your face).
- Read the on-screen script naturally — slight head movement, blink, full range of vowel shapes.
- Wait for processing. Mine took about 90 seconds.
The capture is the only step you can't redo cheaply. Spend 2 minutes here, save 2 hours of failed generations.
Tips for a better Cameo
- Wear a single-color top — patterns confuse the model when it re-renders you in different environments
- Record against a plain wall — busy backgrounds bleed into the output
- Soft, even, front-lit — ring light or a window with diffused daylight is ideal
- Don't smile through the whole script — neutral expressions give the model a wider range
- Re-record if you're not happy with the preview — bad Cameos haunt you forever
Using your Cameo on desktop
Once the Cameo is processed, jump to sora.chatgpt.com on desktop. Your Cameos sync automatically and you'll find them under Drafts → Select. The desktop UI is way better for prompt iteration than the mobile app — you can revise a hook 20 times in 5 minutes.
Creating your first Cameo video (with a viral hook)
The prompt formula that's been working for me:
[Hook line — first 3 seconds] [Visual scene — where you appear] [Action — what you're doing] [Style cue — cinematic / documentary / handheld]
Example: "Wait until you see what happens next. POV: I walk into an Apple Store and every screen shows my face. Cinematic, slow zoom."
The first 3 seconds are everything for retention. If your hook isn't lock-in tight, the rest of the video doesn't matter.
Troubleshooting: content violations & failed generations
Two things will fail your generation:
- Content violations — Sora 2 is strict on celebrity likeness, copyrighted characters, brand logos, and anything sexual/violent. If you're getting rejections, simplify the prompt.
- Cameo drift — sometimes the model loses your likeness mid-scene. Fix: shorter scene, single action, less camera movement.
When in doubt, regenerate with the same prompt. Sora 2 is non-deterministic — you'll often get a usable take on the second try.
The result: a viral intro with perfect lip sync
The final result in the tutorial is a 6-second intro where I appear to be talking to camera in a place I've never been, with audio that sounds like me. From a single 5-second Cameo capture. Total cost: a few cents in Sora credits.
Scale this with the AI Media Machine
Cameos are the shooting step. The bigger problem is scaling — turning one Cameo into 50 hooks per week, scripted, generated, captioned, and queued.
That's what the AI Media Machine does. 12 AI apps in one platform — script writing, hook generation, voiceover cloning, B-roll matching, thumbnail generation, all wired together. Try it for $1.
If you'd rather hand the whole thing off, book a free strategy call and we'll do it for you.
