Wan Image
Generate high-quality images and fix photos instantly with Wan 2.5. The smartest way to create visuals on Somake. No skills required.
After enabling this feature, every image or video you generate will not be automatically shared to the community feed
After enabling this feature, every image or video you generate will not be automatically shared to the community feed
No history found
What is Wan 2.5?
Wan 2.5 is the latest iteration of the foundational generative vision model developed by Alibaba Cloud. This model specializes in both Text-to-Image generation and precise Image Editing (In-painting/Out-painting). Unlike previous generations that struggled with complex semantic instructions, Wan 2.5 offers enhanced prompt adherence, allowing it to understand nuanced descriptions in both English and Chinese.
Key Features
Advanced Semantic Understanding
Wan 2.5 utilizes a large-scale language encoder that significantly improves how the model interprets prompts. It excels at handling long, complex descriptions involving multiple subjects and spatial relationships, reducing the need to regenerate images repeatedly to get the composition right.
Easy Image Editing
This model does more than just create; it fixes. If you have a photo that is almost perfect but has one mistake, you can use Wan 2.5 to change just that part. You can easily remove an object, change a background, or add new items seamlessly.
Enhanced Text Rendering
Wan 2.5 has made great strides in spelling accuracy, allowing it to render short phrases and words much better than older models. However, while the text is legible, it does not always blend perfectly with organic textures.
For example, text generated as a tattoo may look "printed on" or like a sticker sitting on top of the skin, rather than appearing like genuine ink embedded in the body.
Current Limitations
Character Consistency During Editing
When using the editing tools to change a background or environment, you may notice that the character’s appearance shifts slightly. As the AI adjusts the lighting and mood to fit the new scene, the person in the final image might not look 100% identical to the person in the original photo.
Image Authenticity
You may also notice that some images have a distinct "AI look." While Wan 2.5 creates high-definition visuals, they can sometimes appear overly smooth or glossy, lacking the natural imperfections found in real photography. This digital style means the results are often easily identifiable as computer-generated rather than authentic camera photos.
How to Troubleshoot Common Issues?
The text in the image is spelled wrong: Keep the text short (1-3 words). Put the word you want inside quotes, like: "a sign that says 'HELLO'".
The AI ignored part of my prompt: Move the most important details to the very start of your sentence.
The colors look wrong: Try adding specific lighting instructions, such as "cinematic lighting" or "natural daylight," to fix the mood.
Why Choose Somake
Helpful Support
Our team is always available to answer questions and help you improve your prompts.
Simple Editing Tools
Our interface is designed for everyone, making advanced editing as easy as painting with a brush.
Safe and Private
Your uploads and creations are private to you and are not used to train our public models.
FAQ
Yes, you generally own the commercial rights to the images you generate on Somake, so you can use them for ads or products.
Wan 2.5 is excellent at understanding long, detailed instructions and works very well with both English and Chinese text.
Yes, the Somake website is optimized for mobile, so you can generate images on your phone.
