Sora Review: A Landmark Video Model That Is No Longer Available
Sora mattered because it changed what people expected from AI video: longer, more coherent and more cinematic clips than the short, shaky generations that came before it. Its second version added sound and a social app that briefly topped app-store charts. That history is still useful, but it should not be mistaken for a live product, because OpenAI has closed it and no new Sora video can currently be made.
What it is
Sora was OpenAI's family of text-to-video models, first shown in early 2024 and later released as a consumer app built around Sora 2. In simple terms, you described a scene in words, optionally added an image or a reference, and the model returned a short video clip with movement, lighting and camera work. Its later app leaned into social video: people could drop their own likeness into scenes, remix clips and share short, personalised videos in a feed. That combination of a research-grade model and a playful app is why Sora attracted so much attention. It is important to separate the model from the many unrelated websites that now use the Sora name. Only OpenAI's own product was Sora, and that product has been discontinued, so any site still selling Sora access today should be treated with caution.
Key features
The headline feature was coherent, cinematic generation from text. Sora could handle a single continuous shot with camera movement, changes in light and several characters interacting, which made its clips feel closer to filmed footage than to animated slides. It also supported image-to-video, so a still photo could be brought to life, and it could extend a clip forwards or backwards in time. A storyboard workflow let creators arrange several shots into a sequence while trying to keep characters and settings consistent. The Sora 2 app added synchronised sound, so dialogue, footsteps and background noise were generated to match the picture rather than added afterwards. Personalised generation was another distinctive feature, letting users appear inside scenes, which drove much of the app's viral appeal but also raised obvious questions about likeness and consent.
Pricing
There is no current Sora price to recommend, because the product is discontinued. Historically, access was tied to the Sora app and to ChatGPT subscriptions, with a free tier subject to daily limits and paid top-ups for extra generations, while developers could use the API on a per-second basis. Those routes are now closed: the app and website were shut down in April 2026, and the API was switched off in September 2026. Older guides that still list Sora prices are describing a product that can no longer be bought. If you have leftover credits or a subscription question, the relevant route is OpenAI's own support and refund policy, not a third-party reseller. For new work, the practical step is to compare live alternatives on their current pricing pages rather than budgeting around Sora's old rates.
What users like
Users commonly praised Sora's cinematic quality, especially when a simple prompt produced a clip that looked far more expensive than a typical AI generation. Smooth motion was a recurring theme, with walking, camera moves and object movement described as natural for short scenes. The sound in Sora 2 also drew positive attention, because dialogue, footsteps and ambient audio often lined up convincingly with the picture. People who used it for concept work liked how quickly a rough idea could be turned into something watchable, which made it useful for pitches, mood boards and social experiments. Its handling of crowded or atmospheric scenes was another strength mentioned in comparisons, where it often held a scene together better than rivals that produced prettier single frames but weaker movement.
What users complain about
The most consistent complaints were about the details that break realism. Hands were a frequent problem, with fingers merging or extra limbs appearing during complex movement, and faces, clothing and objects sometimes changed subtly as a clip played. Clip length was another limit: even at its best, Sora suited short vignettes rather than full stories, so creators who wanted longer narratives had to stitch many clips together. Some users found generation slow, prompts blocked by strict safety controls, and results inconsistent from one run to the next, so a good first clip did not guarantee a matching second one. Later, the dominant complaint became availability itself: the app closed, the API ended, and people who had built workflows around Sora had to migrate their projects elsewhere.
Limitations to know
Beyond the shutdown, Sora shared the limits of all AI video generation. It could produce warped movement, inconsistent faces or objects across frames, and artefacts in hands, text and fine detail, so any clip needed human checking before use. Its short clip lengths also meant it was never a replacement for filmed storytelling, and its sound, while impressive, could add unrequested background noise or movement that did not match the brief. Personalised video added a further responsibility: generating a recognisable person, especially without their permission, can cause real harm and may breach platform rules or the law, so likeness features should only ever be used with clear consent. Today the most important limitation is simply that Sora cannot be used for new work at all.
Who it is best for
Strictly speaking, Sora is now best for nobody as a production tool, because it cannot generate new video. It remains relevant for researchers, students and creators who want to understand how AI video reached its current quality, and for anyone comparing modern tools against the benchmark Sora set. If you liked what Sora promised — cinematic movement, coherent scenes and matched sound — the practical alternatives are live models with similar strengths, chosen on current pricing, clip length and audio support. Anyone still searching for Sora should first confirm they are not about to pay an unrelated site trading on its name, and should treat any offer of new Sora access as suspect until OpenAI itself announces a successor.
Verdict
Judged purely on what it could do while it was live, Sora was one of the most impressive video models of its generation, with cinematic motion and sound that raised the bar for everyone else. Judged as a tool you can buy and use today, it fails the most basic test, because it is gone. That is why its score here is low despite its historical quality: a review should help you make something now, not admire something you can no longer open. Study Sora to understand the standard, then spend your budget on a live alternative with clear pricing, a working app and support. Our editorial score: 3.0/10 — our own judgement, not an aggregate of user ratings.