Sora 2 gets the press. Wan 2.7 gets the work. Alibaba's video engine has quietly become the model of choice for AI Studio creators making serialized content — short films, branded series, character-driven social — because it solves the problem Sora can't quite crack: holding character identity across multiple shots without drift.
This is a practical breakdown of where each model wins. If you make narrative work — anything where the same person, place or product has to appear more than once — read on.
The continuity problem
Generative video models are trained on individual clips, not on continuity across clips. That training shape produces a known failure: the same character can look subtly different from shot to shot. The face shifts, the wardrobe morphs, the hair changes. On a single hero shot it's invisible. On a five-shot sequence cut together, it screams.
Wan 2.7 was built specifically to solve this. Its multi-shot architecture treats a sequence as one job and locks identity at the model level — not as a post-process anchor.
When Wan 2.7 wins
- Multi-shot scenes with the same character. A conversation, a chase, a journey. Wan holds identity across cuts.
- Serialized social. A weekly character that has to look identical across twelve posts.
- Branded product films. The same SKU across a hero shot, a detail shot and a lifestyle shot.
- Music videos. Performance-driven pieces where the artist appears in multiple looks but has to read as the same person.
Run a five-shot test. Try Wan 2.7's multi-shot mode in AI Studio — write one scene description, get a full sequence with locked identity. Sora 2 lives next to it on the Roster for the moments where it wins.
Download AI Studio ›When Sora 2 wins
- Single-shot organic realism. The found, hand-held, slice-of-life moment.
- Dialogue scenes. Sora's lip-sync is best in class.
- Documentary aesthetic. The piece has to feel filmed-not-made.
- Improvisational tone. The energy is loose, unblocked, alive.
"Sora makes the most realistic single shot. Wan makes the most coherent five-shot scene. They're not the same job." — AI Studio Editorial
Side by side: the same scene, both models
We ran the same five-shot scene through both — a courier picking up a package, walking through a market, delivering it to a customer.
| Beat | Sora 2 | Wan 2.7 |
|---|---|---|
| Pickup | Beautiful, real, organic | Equally beautiful, slightly cleaner light |
| Walk through market | Energy is alive but face shifts subtly | Less energy, identical face |
| Crossing the street | Wardrobe color drifts | Wardrobe identical to beat 1 |
| Approach customer | Looks like a different actor | Same actor |
| Hand-off | Fixable in cut by hiding the face | Cuts clean |
The verdict: for the shoot day energy of a single moment, Sora wins. For the through-line of the scene, Wan wins. The right answer is to use both.
The hybrid workflow
The pros don't choose between Sora and Wan. They cast each engine to the beats it's best at:
- Block the scene. Identify the beats that need organic energy versus the beats that need continuity.
- Cast Sora to the energy beats. Single-shot moments, dialogue close-ups, documentary cutaways.
- Cast Wan to the continuity beats. Anything where the character has to stay consistent across cuts.
- Match grade in post. The two models have slightly different color science. A two-minute pass in your NLE locks them together.
No model switching costs. AI Studio bundles Wan 2.7, Sora 2, Veo 3.1, Kling v3, Seedance 2.0 and the rest into a single iPhone app. Cast each beat to the right engine without juggling subscriptions.
Download AI Studio ›The bottom line
If you're making narrative work, Wan 2.7 deserves a permanent spot in your rotation. It is the model that actually solves the multi-shot continuity problem instead of asking you to work around it. Sora is still the realism king — use it where realism is the whole point. The right tool, on the right beat, wins every time.