OpenAI Sora app amazes in hands-on test after a day’s use

I spent the last 24 hours living in ChatGPT’s new Sora video app — and now I’m a little dazed.

Not only by the fidelity of the clips it makes, but also in how quickly it can turn a selfie and a sentence into something that feels like a film shoot. Sora is much more than another prompt-to-video utility; it’s a fully formed social platform on Android, complete with a TikTok-style feed and a Cameo system that puts your face and voice directly into AI-generated scenes.

What sets Sora apart from other prompt‑to‑video apps

At first, the feed seems familiar: a vertical scroll with likes, comments, and shares. But Sora’s trick is that almost every clip can be reimagined with you in it. The Cameo feature uses your face (via the selfie camera) to scan it in under 10 seconds — faster than when you had to register your mug for a phone unlock, like most devices do these days — and it instantly allows you to create videos in which your likeness appears and speaks with computer-generated audio derived from words typed onto the device screen.

A step up in realism from previous AI video models, Sora is special. Motion follows gravity, shadows fall where you expect them, and camera movements register as intentional rather than janky. Not everything is perfect — anatomy can drift a bit, and hands sometimes look strange — but the overall look is cohesive enough that friends have mistaken my test clips for actual shoots until they spotted the watermark.

Cameo makes you the star with fast face and voice doubles

Once scanned, you can drop yourself into almost any setting: newsroom, ski slope, outer space. Lip‑sync is impressively taut, and skin tone and lighting generally match the scene. You can also share your avatar with friends or others who might want to cast you in your own idea, or make it available widely if you’re an influencer fooling around with scalable presence. The app has even embraced historical figures and public personalities while adding guardrails that have become more restrictive over time, to prevent too obvious abuse.

Sora allows you on iOS to scan objects and animals to bring into the scene with you; your guitar or your cat can co‑star in a drawing created just for fun.

I’ve not yet seen that feature in testing on Android, but the pipeline indicates multi‑subject scenes will soon be commonplace.

How Sora balances video quality, realism, and physics

For now, Sora only makes clips of about a minute in length, which alters what you’re able to attempt. Story beats, scene changes, and dynamic camera work also feel within reach — not just repetitive movement. When testing head‑to‑head with Google’s Veo 3 in tests I did earlier this year, Sora produced more consistently solid physics and fewer visual seams, while Veo could sometimes beat it on crispness of small textures. Veo access also mandates an AI Pro subscription and tends to restrict you to a few eight‑second renders per day, whereas Sora’s app permits up to 30 generations a day free of ads — quite the aggressive play in light of the compute costs modern video models rack up.

If you are coming from image generators like Midjourney, the leap here is coherence over time. Sora also manages to maintain consistent hair, clothing, and lighting as the camera moves around and characters interact — not bulletproof, but certainly a long way from the detached, floaty feel of previous systems.

Availability and access: where Sora works and invite tips

Sora is available to use on mobile in the US, Canada, South Korea, and Japan. Find out how to get an invite if you’re outside those countries while the rollout is ongoing. It’s a generous daily ration of 30 videos to experiment with, and it significantly reduces the barrier for creators trying formats out before they invest in traditional production.

Sora’s social feed is dazzling but still finding focus

As a social network, Sora is still trying to find its soul. Its feed is loaded with eye candy — cats in couture, impossible vistas, wacky comedy sketches — but light on story and human context. And while with Instagram Reels or YouTube Shorts, you tend to stumble on educational explainers or behind‑the‑scenes vignettes, Sora’s clips sometimes feel untethered from reality. It’s transfixing, but minutes in, the novelty can blur into sameness.

With that in mind, Sora’s Cameo opens up a convincing “what if” sandbox: you starring in a mountaineering movie or anchoring a late‑night monologue or performing in the kind of dance routine that you’d never dare to do on camera. That private thrill — creating without sharing — might be the app’s stickiest use case for many.

Privacy, safety, and provenance considerations for Sora

Scanning any platform’s faces and voices should be questioned, after all. Review your settings, restrict the available audience for your avatar, and periodically cull assets you no longer want. OpenAI’s approach accounts for publicly visible watermarks and is “complementary” to overall content authenticity work, like the C2PA framework aimed at tagging synthetic media, a move that industry groups and newsroom standards bodies have called for as deepfakes grow more convincing. Surveys conducted by organizations like Pew Research Center have revealed heightened public anxiety about AI‑created misinformation, and Sora’s palpable definition points visitors in the right direction.

Bottom line: Sora dazzles as a studio, less as a network

As an artificial intelligence video studio in your pocket, Sora is truly impressive. It makes prompts into plausible cinema and makes it possible for you to lead with screenings that take seconds. As a social platform, it’s just less interesting today — beautiful and prolific but often empty. If you’re OK with the privacy trade‑offs and curious what state‑of‑the‑art generative video is capable of today, Sora is a hard app to beat. The results not only exceeded expectations, but also rewired them.