WAN 2.7 brings next-level AI video generation to Picsart

Picsart just made WAN 2.7 its default AI video model, and this one’s worth paying attention to. Sharper visuals, smoother motion, and a 5 reference images and 5 videos capability that no other video model has touched yet. Even better, you can try it for free with up to 5 seconds of 720p video generation on the free tier.

This isn’t a subtle version bump. WAN 2.7 looks noticeably better than its predecessor, with roughly 30% cleaner output across the board. But the real story is control. You can now feed in up to 5 reference images and set both your first and last frame, all at once. That combination? It simply didn’t exist before in any competing model.

Short-form social clips, multi-scene brand campaigns, narrative video projects, WAN 2.7 runs across every Picsart video tool. What follows is a breakdown of what the WAN AI model actually is, where you’ll find WAN 2.7 inside Picsart, and what’s worth trying first.

What is WAN 2.7?

WAN 2.7 comes from Alibaba’s WAN AI video model family. It’s become one of the go-to foundations for AI video generation, and the 2.7 release stretches what the model can do by a wide margin.

The modalities alone tell you a lot: text-to-video, image-to-video, first-to-last frame generation, up to 5 reference images, and 5 videos. That’s an unusually broad toolkit packed into a single model.

Compared to WAN 2.6 (which was already running in Picsart Flow), the improvements are hard to miss. Edges are sharper. Skin tones look more believable. Color balance feels more grounded. Motion got a big upgrade too, subjects move in ways that actually make physical sense, even when the scene is busy or fast-moving. Resolution now tops out at 4K (4096×2160), and you can generate clips at 5, 10, or 15 seconds in all the standard aspect ratios.

Then there’s camera control. You type what you want, “pan left,” “dolly in,” “slow zoom out,” and WAN 2.7 follows the direction. Plain English. No workarounds. It also handles multi-shot video generation with planned transitions and angles, which opens the door for coherent narrative sequences rather than isolated clips. Audio sync improved too. Lip movements, ambient sounds, background music, they actually line up with what’s on screen now. Calling it just a “video generator” undersells it. It’s closer to a full creation toolkit at this point.

What sets WAN 2.7 apart: 5 reference images and 5 videos

This is where things get genuinely different. WAN 2.7 lets you use up to 5 reference images, no other WAN AI model, and no competitor, currently matches that combination.

So what do reference images actually do? They’re visual anchors. Feed in a character’s face, a product shot, a specific environment, brand assets, and the model keeps your video consistent with those inputs. That matters less for a one-off clip and a whole lot more when you’re building something across multiple scenes.

In practice, that means multi-character scenes where every face and outfit holds steady from beginning to end. Product videos where brand colors, packaging, and small details stay accurate. Story-driven content where you cut between shots and the visual continuity doesn’t fall apart. Stack environment, character, and object references together in one generation, anchor the opening and closing frames, and the result is something that actually looks directed rather than randomly generated.

For anyone doing work that depends on consistency across shots, this single feature reshapes the entire process.

Where to find WAN 2.7 in Picsart

WAN 2.7 is live now. It’s not buried in settings or locked behind a toggle, it’s the default. Here’s where each tool fits in.

AI Video Generator

The most straightforward way in. The AI Video Generator runs WAN 2.7 as its default model across mobile native, mobile web, and desktop web. Every modality is available: text-to-video, image-to-video, frame control, reference images, audio. Nothing to configure. You open the tool and WAN 2.7 is already running.

A full-screen generation experience exists on desktop web too, same model, same capabilities, just more screen to work with.

AI Playground

AI Playground is the sandbox. Run WAN 2.7 next to other models, compare how they handle the same prompt, and test edge cases. It’s the right starting point for anyone who wants to explore what different generation modes produce before committing to a workflow.

Picsart Flow

If you were already using WAN 2.6 in Picsart Flow, the transition is seamless. Same workspace, stronger model underneath. Where Flow really shines is chaining steps together, concept to image to video to audio to export, all without leaving the tool. For creators building repeatable content pipelines, WAN 2.7 gives that pipeline a serious quality boost.

Coming soon: Storyline and Aura

WAN 2.7 will roll into Storyline and Aura shortly after launch. Storyline focuses on AI story and animation creation. Aura is more of a creative assistant for image and video editing. Both stand to benefit significantly from what 2.7 brings to the table.

You’ll also see WAN 2.7 surfaced in other spots across the platform, the Featured AI Models card and AI Feed on mobile, the Create Featured banner with a NEW badge on web, and a dedicated WAN model landing page.

What creators can do with WAN 2.7

A few use cases worth highlighting, because the range here is wider than most people expect.

Social content is the obvious one. Type a prompt or drop in an image, and you’ve got a short-form video for ads, product demos, or promos without touching a camera. But the reference image system takes things further. Lock in character faces and outfits through references, and suddenly you can storyboard entire scenes where the same people show up consistently across every clip. That used to require careful manual work or expensive software.

Product photography becomes motion content. A single product shot turns into a lifestyle video, an unboxing sequence, or a styled showcase. Visual explainers and tutorials work well with multi-shot generation, the visuals stay connected scene to scene instead of feeling disjointed.

And if you want the complete pipeline, Flow handles it end to end. Concept to image to video to voiceover, one workspace, no exporting between apps.

WAN 2.7 vs. WAN 2.6: what changed?

WAN 2.7 is a significant step up from 2.6, not just in quality but in what the model actually lets you do. It went from being a video generator to something closer to a video direction tool.

Visual quality: About 30% cleaner overall. Sharper edges, more natural skin tones, more balanced color. You notice it immediately.

Motion: Subjects move the way they should. Fast scenes, crowded frames, things that used to look glitchy now feel grounded.

Reference images: Went from limited support to 5 simultaneous references. That gap is enormous for anyone doing multi-scene work.

Reference videos: Upgraded from 2 to 5 simultaneous references. That gap is enormous for anyone doing multi-scene work.

Frame control: Both opening and closing frames are yours to set now. Version 2.6 only gave you the first frame.

Audio: Voices, music, ambient sound, they sync with what’s actually happening on screen instead of drifting.

Style range: Broader selection, and styles respond to prompts with more accuracy.

Character consistency: Faces, clothing, body proportions hold steady through the entire clip. No more mid-video drift.

Resolution: Jumped from 1080p to 4K (4096×2160).

Camera control: Tell the model what shot you want in plain language. Pan, dolly, zoom. WAN 2.6 couldn’t do this at all.

The short version: WAN 2.6 made video. WAN 2.7 lets you direct it.

Start creating with WAN 2.7

WAN 2.7 is live in Picsart right now. Five reference images, first and last frame control, 4K output, natural language camera direction, tighter audio sync, all of it set as the default across every video tool. No setup, no waiting.

Start creating with WAN 2.7 in Picsart today.