What is Wan 3.0?
“Wan 3.0 lets you create high-resolution, multi-shot videos up to 30 seconds long. Use up to 20 references across text, image, video, audio, and uploaded documents.”
Source: Pika’s Wan 3.0 page. Icon: Pika.
Wan 3.0 is made by Alibaba Cloud, which announced the beta on August 7, 2026. Alibaba says it “supports 30-second video length and multimodal reference inputs,” and that it “can process text, image, video, and audio inputs simultaneously, while also supporting web pages and documents like PDFs and PowerPoint presentations.”
Source: Alibaba Cloud, Alibaba Unveils Wan3.0. Pika’s page lists uploaded documents; check the app for which input types it accepts.
Key features
Up to 30 seconds
Alibaba: “up to 30 seconds per clip allows creators to execute complex camera movements and continuous, unbroken shots.”
Up to 20 references
Across text, image, video, audio and uploaded documents, per Pika.
Faces and voices
Alibaba says it “renders highly realistic human faces with synchronized micro-expressions, produces natural multilingual voice outputs.”
Interfaces and motion graphics
Alibaba highlights “stable software user interfaces and motion graphics.”
Length help and extension
An “intelligent duration feature that recommends the optimal video length,” alongside “video extension features.”
Editing and text
Alibaba’s official repository lists “precision video editing, and advanced text rendering.”
Sources: Alibaba Cloud’s announcement and the official Wan 3.0 repository description. Features described by Alibaba may appear differently in Pika; check the app.
What it costs
Pika’s pricing page lists Wan 3.0 at 33 credits for a 720p, 5-second clip, the lowest of the two video prices it publishes (Seedance 2.5 is 122 for the same clip).
| Plan | Wan 3.0 clips |
|---|---|
| Starter | 27 |
| Creator | 95 |
| Fancy | 259+ |
Source: Pika pricing, checked September 24, 2026. Longer and higher-resolution clips cost more; the app shows the cost before you generate.
A Wan 3.0 workflow
- Open Wan 3.0 and sign in through Pika’s login page.
- Draft the idea at 720p and 5 seconds, the cheapest published option.
- Attach references: images, video, audio or a document, and @mention each one.
- Choose the full length, up to 30 seconds, once the draft works.
- Check the credit cost, then generate.
- Review text, faces and interface elements closely.
- Finish with Extend Video, Color Grade or Pika Soundtrack.
Draft on Wan, finish anywhere. Because Wan 3.0 has the lowest published price, it’s a sensible model for testing prompts. Re-run the winner on Wan at full length, or on another model such as Seedance 2.5 for comparison.
References and documents
Wan 3.0’s stand-out input is documents. Instead of describing data, you can attach it.
| Reference | Use it for |
|---|---|
| Text | Scripts, briefs, voiceover lines |
| Image | Characters, products, locations, style |
| Video | Motion, camera movement, pacing |
| Audio | Music, voice, sound mood |
| Document | Reports, slides and data to turn into explainers |
- Pick the key points in your prompt; don’t expect every line of a long document to appear.
- Check every number and label in the video against the document.
- Only upload documents you’re allowed to share; avoid confidential material.
Prompt tips for Wan 3.0
- For long takes, describe the camera path from start to finish: “starts on…, rises past…, ends on…”.
- For multi-shot videos, number the shots and give each one framing and action.
- For explainers, list the sections you want: title, three points, summary.
- For dialogue, put lines in quotes and name the language.
- For on-screen text, keep it short and in quotes.
- Use @mentions for every reference, with a “keep … exactly” line.
Example prompts and results
Six prompts covering drafts, long takes, documents, references, interface text and multilingual dialogue. Turn your sound on. The prompts are our own.
Cheap 5-second draft
A quiet fishing harbour just after sunrise. Wooden boats rock gently, mist over the far shore, gulls drift across the frame. Slow lateral tracking shot, 35mm look.
What to look for: Whether the idea works before you spend more credits on a longer version.
Try one change: Once it works, re-run the same prompt at a longer length or on another model.
One continuous 30-second shot
One unbroken shot: the camera starts on a street café table, rises past the awning, glides along the building facade and ends high above the city at dusk as the lights come on. Smooth, steady movement, ambient city sound.
What to look for: A single continuous move with no cuts, holding steady for the whole length.
Try one change: Split it into two shots instead and compare.
Explainer from a document
Turn @report into a 20-second animated explainer: title card, three key numbers shown as simple motion graphics, and a closing summary. Clean flat style, calm narration.
What to look for: The numbers and labels match the document exactly.
Try one change: Ask for a vertical 9:16 version for social media.
Product with image and audio references
@bottle on a wet stone surface as water splashes around it in slow motion, then a slow orbit. Use @track for the music. Keep @bottle’s label and colour exactly as in the reference.
What to look for: The label stays sharp and the music follows the reference track.
Try one change: Swap the music reference for a different mood.
App demo with on-screen text
A phone screen showing a simple to-do app. A finger taps “Add task”, types “Buy flowers”, and the item slides into the list with a tick animation. Clean UI, soft background, gentle click sounds.
What to look for: Interface text stays stable and correctly spelled as it animates.
Try one change: Change the app’s colour theme and keep everything else.
Character speaking two languages
@host stands in a bright kitchen and says, “Welcome to the show,” then repeats it in Spanish: “Bienvenidos al programa.” Medium shot, natural light, warm tone. Keep @host exactly as in the reference.
What to look for: Natural voice in both languages and facial expressions that match the words.
Try one change: Try a third language and check the lip movement.
When to choose Wan 3.0
Choose Wan 3.0 for
- Cheap drafts and prompt testing
- Long, continuous shots up to 30 seconds
- Explainers built from documents
- App demos and motion graphics
- Many references of different types
Consider another model for
- Up to 50 references and 4K export: Seedance 2.5
- Native stereo dialogue in short scenes: MiniMax H3
- Keyframe transitions: Pika 2.5
Compare all models in the video models guide.
Wan 3.0 FAQs
How long can Wan 3.0 videos be?
Up to 30 seconds, according to Pika and Alibaba.
How much does Wan 3.0 cost in Pika?
Pika lists 33 credits for a 720p, 5-second clip; longer or higher-resolution clips cost more.
Can I make a video from a PDF?
Pika lists uploaded documents among Wan 3.0’s references, and Alibaba mentions PDFs and PowerPoint presentations. Check which file types the Pika app accepts.
Does Wan 3.0 make sound?
Pika’s page doesn’t mention audio output, but Alibaba says Wan 3.0 “produces natural multilingual voice outputs.” Check the options in the app, or add sound with Pika Soundtrack.
Who makes Wan 3.0?
Alibaba Cloud. Pika offers it alongside its own Pika 2.5 model.
Is this the official Pika or Alibaba website?
No. pikaais.com is an independent informational guide, not affiliated with Pika or Alibaba.
Draft first, then go long
Copy prompt 01 and generate a 5-second draft for 33 credits. If it works, try prompt 02 as one continuous 30-second shot.
Open Wan 3.0 ↗Next: Seedance 2.5 · Video models · AI video hub
Sources: Pika Wan 3.0, Pika pricing, Alibaba Cloud’s Wan3.0 announcement and the official Wan 3.0 repository, checked September 24, 2026. Images and icon: Pika. Prompts and tips are our own. Not affiliated with Alibaba.