Tongyi Wanxiang

Tongyi Wanxiang is the visual generation product from Alibaba’s Tongyi team, a sibling of the Qwen chatbot. Its web app covers both images and video: write a description and get pictures, or turn text or a still image into a few seconds of footage. Alibaba has also open-sourced some of the Wan video models behind it, which gives it more weight among developers than a typical consumer app.

For users in mainland China its biggest advantage is that nothing needs setting up: it is directly reachable, you sign up with a phone number, and it reads Chinese prompts well. Descriptions rich in cultural context — a rainy Jiangnan alley, flagstones, a woman with an oil-paper umbrella — land more faithfully here than when handed to an overseas model.

Think of it as a big platform’s general-purpose creative entry point rather than a specialist tool chasing a signature look. Its output is among the best of the Chinese products, but its aesthetic ceiling and detail consistency still trail the top overseas services. It works well as the workhorse for everyday illustrations and video drafts, less well as the only source for final deliverables.

Key features

• Text-to-image: Generate images from Chinese or English prompts, with aspect ratio and preset styles, returning several candidates at once.
• Reference-based generation: Upload an image and get new ones with a similar style or composition — handy for a consistent series.
• Sketch to image: Scribble a rough drawing, add a short description, and the model fills it in. Friendly to people who don’t write long prompts.
• Text-to-video: Generate a short clip straight from a description, steering camera movement and mood through the prompt.
• Image-to-video: Add a motion description to a still and bring it to life — often used for animated posters or social clips.

How to use

1. Open the Wanxiang web app and sign in with an Alibaba Cloud or Tongyi account; a phone number is enough to register.
2. Switch to image or video generation and pick a mode such as text-to-image, reference image or image-to-video.
3. Describe subject, setting, style and lighting, choose a ratio and submit. Video jobs may queue at busy times.
4. Pick the results you like, adjust the prompt or reference if needed, then download.

Best for

• Blog and social illustrations: fast, good with Chinese context, and more than adequate for daily use.
• Short-video material and animated posters: image-to-video turns existing assets into motion quickly.
• Chinese cultural subjects: festivals, classical styles and regional scenes come out more on-topic than with overseas models.
• Teams with restricted networks: directly reachable in China with no extra setup, so easy to roll out internally.

Strengths and limitations

Strengths

• Reachable in mainland China, easy sign-up, accurate understanding of Chinese prompts.
• Image and video in one place, so going from still to motion needs no tool switch.
• An open-source model ecosystem behind it if you later want self-hosting or API access.

Limitations

• Aesthetic ceiling and detail stability still lag the best overseas tools; hands and in-image text sometimes go wrong.
• Content moderation is strict, and ordinary creative prompts are occasionally blocked until rephrased.
• Video generation can queue for a long time at peak hours, and slows noticeably once free usage runs out.
• Commercial rights depend on the user agreement, and generated content may carry an AI label as required by regulation; check before commercial use.