Wan 2.5 Preview Released: Top 6 Big New Features for AI Video & Image

Alibaba has officially launched its next-generation AI Model, WAN 2.5 Preview. This release marks a significant step forward for AI in video and image generation, with its new architecture and powerful features set to revolutionize how we create and edit visual content.

Imagine crafting a cinematic short film with perfect music, sound effects, and character dialogue, all with just a few simple commands. This is no longer science fiction; it's the reality brought to you by WAN 2.5 Preview. Best of all, you don't need any special equipment or complex setups. The Model offers a fast online experience that brings your creative ideas to life instantly.

Top 6 Big New Features in Wan 2.5 Preview

Here is a quick look at the biggest new features that make Wan 2.5 Preview a major step up over previous versions:

1. Native audio-visual sync — Dialogue, ASMR, music, and sound effects now align perfectly with the visuals, killing the classic "out-of-sync" problem in AI video.
2. Up to 10-second video clips — Create longer, richer stories instead of short looping animations.
3. Up to 1080p cinematic quality — Higher resolution output with stronger motion and structural stability.
4. Audio as an input — Generate video that matches the mood of a song, voiceover, or sound effect you provide.
5. Conversational image editing — Change colors, objects, or backgrounds with simple instructions like "make the car blue" or "swap the background to a snowy mountain."
6. Enhanced text-to-image & prompt following — Photorealistic images, artistic styles, and charts that closely match your intent.

Wan 2.1 vs Wan 2.5: Image and Video Feature Comparison

For creators searching for Wan 2.5 image capabilities, the biggest change is that Wan 2.5 brings image generation and conversational image editing into the same multimodal creative workflow while also improving video quality and duration.

In short, Wan 2.1 remains a strong choice for open-source and local workflows, while Wan 2.5 Preview is aimed at creators who want higher-resolution output, longer clips, native sound, and more capable image creation and editing.

Core Breakthroughs: The Magic of Audio-Visual Fusion

The most remarkable innovation in WAN 2.5 Preview is its ability to overcome the common "out-of-sync" problem in AI-generated video, achieving perfect audio-visual synchronization.

• Sound and Picture in Perfect Harmony: Previous AI videos often had mismatched audio, with characters' mouth movements not aligning with their speech, or background music failing to match the scene's mood. WAN 2.5 Preview solves this from the ground up. Whether it’s character dialogue, ambient sounds (like ASMR), or background music, everything aligns precisely with the visuals, making your creations feel more authentic.
• AI Filmmaking: The Model can generate 10-second 1080p HD videos with cinematic quality, boasting impressive dynamics and structural stability. This means you’re no longer limited to short, looping animations; you can use AI to tell richer, more compelling stories, unlocking a world of creative possibilities.
• Audio as a "Paintbrush": Creators can even use audio as an input source. For instance, you could provide a piece of soothing music or a specific sound effect, and the AI will generate a video that matches the audio's atmosphere. This truly enables a "painting with sound" approach.

How to Experience: All these powerful video features are available on the WAN 2.5 Preview online Model. Just visit the website, enter text or upload audio, and effortlessly generate your own cinematic videos without worrying about technical configurations.

Try it Now

Smart and Precise: Taking Image Creation to the Next Level

Beyond its impressive video capabilities, WAN 2.5 Preview also offers a huge leap in image creation.

• Understanding Your Vision: Past AI image generators often produced strange results from simple prompts. WAN 2.5 Preview's instruction-following ability is greatly enhanced, allowing it to more accurately fulfill your requests. Whether you want photorealistic images, diverse artistic styles, or professional-grade charts, it produces results that closely match your intent.
• "Magical" Editing Functions: Its image editing feels like magic. With simple conversational commands, you can precisely edit images. For example, you can say "change the car's color to blue" or "replace the background with a snowy mountain," and the AI will execute these complex tasks with perfect, pixel-level precision.

How to Experience: Log in to the Model now, upload an image, and try editing it with simple, conversational commands. You'll be amazed by the AI's understanding and its ability to deliver precise, pixel-perfect results.

Try it Now

Showcase

A big leap forward for image & video generation

• Audio-Visual Sync (voices, ASMR, music, effects)
• 10s Videos (richer storytelling)
• 1080P cinematic quality
• Audio conditioning as input

Generation with music

Character with voice

Sound Effect

The Technical Foundation: A Unified Multimodal Brain

These groundbreaking features are all thanks to WAN 2.5 Preview's new unified multimodal architecture. This is like giving the AI a single brain that can simultaneously see, hear, read, and write, integrating the processing of text, images, video, and audio into one seamless framework. Through joint multimodal training, the model achieves stronger alignment between different data types, which is crucial for perfect audio-visual sync and precise instruction-following.

In short, WAN 2.5 Preview is more than just a tool; it's an intelligent creative partner. It makes professional-level video and image creation incredibly simple, putting the power of creativity back into the hands of everyone.

Wan 2.5 Preview FAQ

When is the Wan 2.5 release date?

Wan 2.5 Preview became available in September 2025. It is currently in preview mode, and Alibaba has not confirmed a separate final-release date yet. We will update this article as soon as an official date is announced.

What are the biggest new features in Wan 2.5?

The biggest new features are native audio-visual synchronization, 10-second 1080p video generation, audio conditioning as an input, and conversational image editing. Together, they let you create cinematic videos with dialogue, music, and sound effects directly in your browser.

How is Wan 2.5 different from Wan 2.1?

Wan 2.5 Preview adds native audio generation, audio-input video generation, conversational image editing, longer clips (up to 10 seconds), and up to 1080p resolution. Wan 2.1 remains a strong choice for open-source, local, and offline workflows.

Can I generate videos with audio in Wan 2.5?

Yes. Wan 2.5 Preview can generate synchronized dialogue, music, ambience, and sound effects alongside your video. You can also provide audio as an input, and the model will generate a video that matches its mood and atmosphere.

Can Wan 2.5 edit images?

Yes. Wan 2.5 Preview supports conversational image editing. You can give simple instructions — such as "change the car's color to blue" or "replace the background with a snowy mountain" — and the AI will apply precise, pixel-level edits.

Does Wan 2.5 support image-to-video?

Yes. Wan 2.5 Preview supports image-to-video generation with improved image understanding and visual consistency compared with Wan 2.1, so your generated clips stay faithful to the reference image.

Is Wan 2.5 free to use?

Wan 2.5 Preview is available through an online preview that requires no installation. Visit the Wan AI website, enter your text prompt or upload an audio/image, and start generating — no technical setup needed.

Do I need powerful hardware to run Wan 2.5?

No. Unlike Wan 2.1's local deployment option, Wan 2.5 Preview runs online in the browser, so you don't need a high-end GPU or any special equipment to use its features.

What video resolution and length does Wan 2.5 support?

Wan 2.5 Preview supports videos up to 10 seconds long and up to 1080p resolution with cinematic quality and stable motion.