# Dreamerland All in one AI art generator Dreamerland is an advanced AI art generator that empowers users to create unique and breathtaking pieces of art using the power of artificial intelligence. In this tutorial, we will guide you through the steps on how to effectively utilize Dreamerland and unlock your creative potential. 1. **Choose the model you want** We provide several models for you to choose from. Different models are fine-tuned for different purposes. Some models are designed for generating Anime art, some for generating artistic images, while others are specifically tailored for creating photorealistic images. ![](https://2310956826-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Ftbrc6W2n35oVxftMovDQ%2Fuploads%2FpA0dtQVZdd31zqyrN7Br%2FIMG_7772.jpg?alt=media\&token=6a4fa71f-b9be-4021-af82-796453fd53fb)![](https://2310956826-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Ftbrc6W2n35oVxftMovDQ%2Fuploads%2FOVFG3W77XgFhh8IhxERN%2FIMG_7771.jpg?alt=media\&token=68022701-8715-44d5-b031-fc43d23be0fb) **Anime:** Anime Diffusion V2, Anime Diffusion, Meina's, Dark Sushi Mix, RealCartoon3D, Galaena REDUX, Mistoon Anime **Photoreal:** Dreamer Photoreal V2, Dreamer Photoreal, majicMIX realistic, Realistic Vision, XXMix 9realistic, epicRealism **Artistic/General purpose:** Dreamerland, DreamShaper 2. **Write your prompt** Your prompt is where you will begin writing about what you dream of seeing in the art you want to generate. Please describe your idea clearly. The more descriptive you are in your prompt, the more vibrant and unique the output will be. The sequence of words matters. Words near the front of a prompt are weighted more heavily than the words at the back of a prompt. You can usually follow this order to write your prompt: 1. Start with the subject and art style. 2. Provide a detailed description of the subject. 3. Describe the background, including the location and time. 4. Specify the lighting and angle. 5. Include any additional details you would like to add. To adjust the weights of words, you can increase the weighting by using parentheses "()," or decrease the weighting by using square brackets "\[]". You can even use multiple parentheses like (((white hair))) to further emphasize certain words.

3. **Select preset styles you want to add into your art (optional)** We offer a wide selection of hundreds of preset styles for you to choose from. You have the freedom to mix and match different styles to create your unique art. You can select up to a maximum of 12 styles to incorporate into your artwork.

You can click on the selected style to adjust its weight, which will affect how much influence the style has on your art.

4. **Negative prompt and negative embedding** The inclusion of a negative prompt is crucial. Please specify something you do not want to appear in your art. Providing a precise negative prompt will significantly enhance the quality of the output. Additionally, you have the option to choose from our preset negative embeddings. This can be helpful if you're unsure about what to write in the negative prompt, as it provides a simple and convenient alternative. For further information on negative embeddings, please refer to the following resource:[Negative embeddings](/basic-terms/negative-embeddings) 5. **Other settings** We provide users with a great deal of freedom to customize their experience. You can adjust the settings according to your specific needs. In another section, we will explain the various settings and how they work in detail. 6. **Start Drawing Now!** Click the yellow 'Start Drawing' button to begin. You are now ready to create your first AI art masterpiece. # Hires. fix High Resolution Fix - Upscale the image and fix details High-Resolution Fix (Hires. Fix) is a process that involves upscaling an image and subsequently redrawing it based on the original image. Hires Fix is essential for achieving high-quality images as it can enhance the characters' faces and add additional details. However, it is worth noting that this process may alter the images to some degree, depending on the strength value applied. **How does it work?** We create a image with prompt "a woman wearing tee shirt walking on a street, full body". The resulting image is in a good composition but the face of the woman is obviously deformed.

We can use the 'High-res Fix' feature, which allows us to upscale and enhance the image, focusing specifically on improving the face. To access this feature, simply click on the 'High-res Fix' button located at the bottom of the page. It will provide you with a range of settings to adjust for the high-resolution fix.

If you are unsure about the various settings, you can leave them at their default values and simply click the 'Fix Now' button to initiate the high-resolution fix process.

After the high-resolution fix process, you will notice a significant enhancement in the face, as well as improved details on the hands and shoes. The overall image quality is greatly improved, and it is upscaled to 2X its original size. # Hires. fix with different model High resolution fix in different styles Building upon the previous example, during the high-resolution fix process, you have the option to utilize different models. By selecting different models, you can achieve distinct styles in the final output, providing you with a range of possibilities to explore.

We changed the model from Photoreal V2 to Anime V2

When you look at the final image, you will see that it has transformed into an anime style while keeping the original composition and details intact. The high-resolution fix process has also corrected and improved the appearance of the face and hands.

# Optical illusion A image appear as one thing from a distance, but up close they reveal something completely different **What is optical illusion?** Optical illusion is an illusion caused by the visual system and characterized by a visual percept that arguably appears to differ from reality. **How does it work?** Here is a demonstration on how to generate an image that makes a landscape look like a woman. 1. Go to the shortcut page and click on "Optical illusion." 2. Select a reference image. 3. Write down the prompt for the illusion. We will use a simple prompt such as "river, sky, forest." 4. You can click on the control layer to adjust the conditioning scale. The higher the value, the more the resulting image will follow the reference image, but it will be less creative.

After everything is set, click the button to start generating. Wow, the result is amazing. Can you see the woman in the landscape picture?

# Inpainting Inpainting is a technique used to modify or restore masked areas in an image. Depending on the strength of inpainting, it can serve two purposes. Firstly, it can be employed to create entirely new content within any region of an existing picture. Secondly, it can be used to subtly restore deformed parts of an image. Let's look at a simple example of inpainting. We will use this image as a base image.

A girl with black hair, wearing school uniform, in classroom

We will try to change her hair color by inpainting function. Follow the steps below: 1. Choose Inpaint Mode 2. Select a source image from your phone 3. Mask the hair area in the image 4. Enter "1girl, pink hair" in prompt

Now we can hit Start drawing button to see the result. The final output is absolutely awesome.

# BrushNet BrushNet is a novel image inpainting model hat focuses on preserving image details and structures while filling in missing or masked regions. Unlike traditional inpainting methods that may distort original content, BrushNet ensures high-fidelity reconstruction by incorporating multi-stage fusion and reference feature injection. Version 1.34.1 introduces the BrushNet option for inpainting, **currently available exclusively for SDXL models**. **For subtle image edits, it's better to use low-strength inpainting (around 0.4-0.6) without BrushNet, as BrushNet tends to make more noticeable changes to the inpainted area.**

Check the box next to BrushNet to enable this BrushNet feature.

Let’s take a look at the examples below to see the difference between inpainting with and without BrushNet. In this example, we’ll add a beer to the table in the photo by masking the region for inpainting.

Let's see the result with and without BrushNet:

The results clearly show that BrushNet preserves the original image structure while seamlessly adding the beer to the table. # Inpaint not masked We have introduced the fundamental inpaint function, and now we will discuss the concept of "inpaint not masked" in this section. There are instances when you may want to preserve a specific small area, typically a face, while redrawing or modifying the rest of the image. Let's see how this work with a simple example. We will use this image as base image in this section

a young girl, wearing schcool uniform, in classroom

Similar to the normal inpainting process, we begin by selecting the inpaint mode and proceed to draw a mask. However, in this case, we draw the mask specifically on the area that we wish to preserve.

After drawing the mask, remember to check the "Inpaint not masked" option. To maintain accurate body composition, it is advisable to utilize a control layer in conjunction with the inpainting process. In this case, we will employ Pose control with the base image.

Check "inpaint not masked", employ Pose control layer

Lastly, input the desired prompt. In this case, let's use the following prompt: "1 girl, wearing bodysuit, pink hair, long hair, in cyberpunk city." Now, click on the "Start drawing" button. Below is the resulting image we obtained:

# Regional Prompt (1.40.0) The regional prompt has been upgraded in version 1.40.0. This new regional prompt is more powerful and easier to use, enhancing the user experience significantly. ### **What's the update** 1. **Base prompt** * The base prompt is now consistently applied across all regions of the image, ensuring uniformity and coherence in output. 2. **Isolate Lora style by region** * You can now use Lora style specific to each region without interference. This resolves the previous issue where styles would affect other areas unintentionally, providing more control over regional styling. 3. **Color map grid editor** * You can utilize the color map grid editor to customize your color maps effortlessly. This tool allows you to modify colors and patterns without starting from scratch, ensuring efficiency and precision in your design process. ### How It Works #### Example: Creating Three Anime Characters in One Image To create a cohesive image featuring three distinct anime characters using the styles and color map grid editor: #### Step-by-Step Guide 1. **Create a Color Map**: Utilize the grid color editor to create a color map. Select three equal columns and click "Generate Color Map" to establish the color palette for your image.

2. **Write a Base Prompt**: Next, we'll need to draft a base prompt for the entire image as base concept. Given that we have three girls as characters in the image, a suitable prompt might be: "3girls, standing in a living room."

3. **Craft Area-specific Prompts**: Write prompts and apply styles for each area. You should write a prompt specifically for each distinct area.

4. Hit the create button to check the result.

5. Bonus You can use the control layer to adjust the poses of all three characters in the image.

Here is the final result:

# Regional Prompt (color mask) Control image composition in one generation **What is regional prompt?** It is a tool that allows you to specify different prompts for different regions. **Why use a regional prompt?** Have you ever found it difficult to create multiple subjects in a single image? Here is an example:

try to generate a image with 2 girls. one girl with school uniform on the left. another one with bikini on the left

You will probably get the result below with the above prompt. The resulting image is definitely not what we are thinking of.

Regional prompt is your helper. We can specify different prompts for the girl on the left and the girl on the right using the regional prompt. **How does it work?** Enable the regional prompt feature with the toggle. It will take you to a color map page.

Draw a color map for image composition. For example, we want a girl with a school uniform on the left in the red area, a girl with a bikini on the right in the yellow area, and a starry sky on top in the white area. Here is the color map we drew:

Great! We have a color map now. Let's create a prompt for each area. We enter the prompt for each area as shown below. \**Remember to specify "2 girls" in the prompt; otherwise, it will become "1 girl" wearing half a school uniform and half a bikini.*

Start generating! We have a great image with the correct composition!

# Regional Prompt with Reference Image Control the image composition with control layer You may want to have more control while using a regional prompt. Here, we bring you a demonstration of using a reference image to draw a color map and using the reference image in the control layer. **How does it work?** In this example, we want to generate an image of a woman with half white hair and half black hair. First, open the color map drawing canvas. Click the "Background" button to select a reference image.

Now, let's draw a color map on the reference image. We want the woman to have half white hair on the left and half black hair on the right. We'll use the color red for the left part of her hair and yellow for the right part.

Here is a final color map we have. Next step is to write prompts for each area.

Don't forget to add a control layer with the reference image. We used a depth control here.

Okay, now we can click a button to generate images! Below are the resulting images of this configuration in both the realistic model and anime model.

# Regional Prompt (prompt mode) **Why use a regional prompt?** Have you ever found prompt bleeding in your image (one part of prompt effecting another part)? Let's see the below example:

We want to create a image with a Spanish woman wearing a forest printed tee shirt with Japanese temple around. We get the results below with this prompt.

The tee shirt print is influenced by "Japanese temple". The skirt is affected by "printed" to be a printed skirt. Additionally, even the woman's face is affected by "Japanese" and "Japan" and she looks like an asian woman. The regional prompt is designed to prevent prompt bleeding. **How it works?** Let's try to move some descriptive words to regional prompt like this.

Below are the results of the above setting

The result appears to be perfect, which precisely aligns with what we aimed for in our prompt. # Outpainting # Image prompt An image as part of the prompt to influence the output image [**What is image prompt?**](#user-content-fn-1)[^1] Image prompting is a technique used in AI image generation where an image is provided as an additional input to the text prompt. This image influences the composition, style, and colors of the output image. Both the text prompt and the image prompt act as conditioning factors, guiding the AI in generating an image that aligns with the desired visual elements and characteristics. By incorporating an image prompt, users can exert more control over the creative process and achieve specific artistic effects or styles in the generated images. **How it works** We would like to create an image featuring a golden chair and a golden table. To begin, we select a model and input a text prompt.

Let's observe how it looks without an image prompt.

Next, we incorporate an image prompt into our generation process. We have discovered an image featuring gold that will serve as our prompt. In this instance, we only desire the gold color and texture, so we set the strength to 0.5 to prevent the AI from extracting the gold's shape.

Once more, we click the "Start drawing" button and see the final result. That's perfect.

Read more about image prompt on {% content-ref url="/pages/xAfo7n5yT5ibObFFdPKQ" %} [Consistent face](/image/image-prompt/consistent-face) {% endcontent-ref %} {% content-ref url="/pages/375lgr6Sv9qnwIUYM8g4" %} [Consistent anime character](/image/image-prompt/consistent-anime-character) {% endcontent-ref %} {% content-ref url="/pages/YQK7m7qdzXZG8w6Wll12" %} [Style](/image/image-prompt/style) {% endcontent-ref %} [^1]: # Consistent face We can achieve consistent faces using an image prompt with just one image. Let's take a look at this example. (Please be aware that this feature does not involve face swapping. It may look similar but It will not replicate the face exactly as it appears in your input image.)

Below are the results we can obtain with various text prompts.

a woman, wearing business suit, in office

# Consistent anime character

Below are the results we can obtain with various text prompts using the image prompt setting above.

1boy, wearing school uniform, in classroom

1boy, topless, pants, on beach, sunlight

# Style We can apply different styles to the image with image prompts. Let's draw a anime girl with control layer.

Next, while keeping all the settings the same, we apply different image prompts to see the differences.

# Doodle Doodle, a captivating feature that unleashes your creativity, allows you to draw and transform your sketches into magnificent works of art. With Doodle, the possibilities are endless as you effortlessly wield virtual brushes, pencils, and colors to bring your imagination to life.

Below is the result of our drawing Remember to type prompt to describe your image: 1girl, red hair, wearing bikini, grass, blue sky, red moon, holding a sword

You can also add a image to your drawing like this example below. We add a sniper gun image to our drawing. Click the +Add button to add any image assets you want.

1girl, holding a sniper gun, in swimming pool, wearing bikini, purple hair

Doodle empowers you to express yourself visually and share your creations with the world. # Inpaint with doodle We have introduce a doodle feature. You can draw the image you want before inpainting. Take a look at below example. 1. go to inpaint page, we can click "Edit" to draw on the image

2. Draw on the image. In this example, we want to draw a print on her tee shirt.

3. Mask the inpainting area that we want to modify. We should set the strength to something around 0.5 so it will follow our drawing.

This is the output of our work:

# Area inpaint with doodle For smaller area, we can go for area inpaint. It will give you a better result than normal inpaint. 1. Select area we want to inpaint

2. Click on the Doodle button. It will direct you to the drawing page

3. Draw on the area. In this example, we want add a tattoo.

4. Remember to edit the prompt to match our drawing. We add tattoo text "dream", tattoo red heart to the prompt. We should also adjust the strength to around 0.5. It will follow what we draw.

It is recommended to perform a hires fix after that. Here is the result of our work.

# Relighting Adjust the lighting of an image to create a more dynamic and visually appealing effect.

This feature allows you to change the direction of the light source in an image. You can relight your image or upload a new one.

Next, enter prompt of the light. For examples, sunset / neon light / soft lighting / dark atmosphere / spot light etc Adjust the light source direction:

Adjust the effect weighting:

Hi the "Start now" button and check the results

You can also relight anime image:

# Face Swap Apply a facial feature from one image to another This innovative tool allows users to seamlessly exchange faces in photos with just a few taps. First, you will see a face swap buttons at the bottom in your artwork page

Upload a face image and select the face that you want to apply. Then click "Swap now"

These are the result images by applying different faces.

# 3D Parallax effect Turn 2D images into 3D parallax effect video Enhance your images with stunning 3D parallax effects! Follow these simple steps: ### Where to Find 3D Parallax * Navigate to the **Tools Page** or scroll to the bottom of the **Image Page**. * Choose an image from your device. * Alternatively, use an image generated within the app.

### Customizing Your Parallax Effect 1. **Select Camera Motion**: * Pick the camera motion style that appeals to you. 2. **Adjust Parameters**: * **Motion Strength**: Set how pronounced the effect will be. * **FPS (Frames Per Second)**: Adjust the speed of the motion. * **Loop**: Choose whether the effect should repeat.

### Generating Your 3D Parallax * Once you’ve set your preferences, click the **Start Now** button to generate your 3D parallax effect!

# In-context transfer In-Context Transfer allows users to edit images intelligently by leveraging the context of existing visual elements. By using a reference image, users can seamlessly integrate new elements or modify existing ones, ensuring that changes are harmoniously blended into the overall composition. ### **How to use**

* **Go to Flux model and select In-context transfer mode** * **Upload Your Source Image**: Start by selecting the image you’d like to modify. * **Choose a Context Image**: Select an image that represents the style or elements you wish to incorporate. * **Draw the Inpainting Area**: Highlight the section of your source image where you want to apply the inpainting. * **Draw the area that you want to transfer:** Highlight the section of your context image where you want to transfer to the source image * **Enter the prompt to describe your desired output, it also work if you leave it blank**

Hit the Start Drawing button and wait for the result:

### Other examples

It also work on anime image

You can transfer anything

# Elements We’re thrilled to introduce **Elements**, our latest feature that empowers you to create truly unique visuals by blending and customizing different image elements. This feature gives you the freedom to mix, match, and modify various elements—colors, textures, shapes, and more—to craft images that perfectly match your vision. With **Elements**, you can: * **Combine multiple image elements** to build dynamic compositions. * **Adjust styles, effects, and layouts** for a personalized touch. ### **Subject** Generate high-fidelity images of characters, objects, or animals while preserving key details. Perfect for consistent branding, character design, and product visuals, this feature ensures accuracy and identity retention.

A woman playing guitar on stage, neon text "Dreamer" in background

### Face This mode focuses solely on facial feature. Perfect for consistent face.

portrait, made of sculpture, white color

### Combine elements Combine multiple elements image (maximum 3) into one image.

a woman holding a doll with both hands in living room

# Qwen Image Edit Qwen Image Edit extends the capabilities of Qwen-Image into the realm of image editing.You no longer need to draw a mask to inpaint an image. Simply provide an instruction, and your image will be ready. #### Key Features: * **Semantic and Appearance Editing**: Offers both detailed visual edits—like adding, removing, or modifying specific elements while leaving other areas untouched—and broader semantic changes such as intellectual property creation, object rotation, and style transfer, ensuring semantic consistency even when pixels are altered. * **Precise Text Editing**: Supports precise text manipulations in both Chinese and English, enabling direct text additions, deletions, and modifications while preserving the original font, size, and style of the image. ### Get start

			Cover image
Single element	Edit 1 image with a prompt	Qwen Image Edit	wEYoEM6eDY32VYhKnUvR_0.png	Qwen Image Edit
Multiple elements	Adding more than 1 image	Multiple elements	result1.jpg	Multiple elements
Blending elements	Placing elements on base image	Blending elements	a5bjELqILBZyskdwSeh5_0.png	Blending elements
Generative fill	Fill the missing part	Generative fill	dl_1756841037283.jpg

### How it works 1. Upload a image that you want to edit 2. Enter a instruction prompt 3. Select styles (optional) 4. Wait for the output Let's walk through an example: 1. First we input this image for editing

2. Type the prompt as instruction

Below are the results of different prompt

3. We can also select a community style for different concept For example, we select the figure maker style. We used the default trigger prompt. Here is the result.

Editing an image has never been easier. The result is impressive. # Multiple elements You can add more than 1 element in Qwen Image Edit. ### How it works 1. Add two or three images in elements.

2. Write a instruction prompt to tell AI how to combine two images. Tips of writing prompt: * You should tell the AI what subject you are referring to. In this example, the girl is on image 1 and the sword is on image 2. so we write "make the girl on image 1 holding the sword on image 2". If you have three elements, you can specify the subject on the image 1, image 2, or image 3. 3. That's it. Hit the Draw button and wait for the output. Below is the final output and it looks perfect

# Blending elements Another usage is to use Lora to blend different elements. For better consistency, you can use "Put it here" Lora to do this.

### How it works 1. Upload a base image and click the edit button on the bottom left

2. Click the "Add Image" button to add elements on top of the base image. Position the elements where you want them to be applied. For example, we want to change her clothes to a bikini, so we place the bikini image on her body.

3. Select "Put it here" style and write a instruction prompt. Tips of writing prompt: 1. You should tell the AI what to do. In this example, we want to remove the white area of the bikini image we want to remove her clothes we want to replace her clothes to this flag bikini\ so the prompt is "remove white area, remove her clothes, change her clothes to flag bikini" 2. Describe the final output. In this example, The final output is a girl wearing bikini 3. The "Put it here" style is to maintain consistency in image editing, although it is optional. 4. That's it. Hit the Draw button and wait for the output. Below is the final output and it looks perfect. Details of the bikini and person remain consistent.

Here are output of with and without "Put it here" Lora. You can see the difference in consistency of the bikini.

# Generative fill You can inpaint and outpaint with Generative fill lora. It will fill the missing part of the image. ### Inpaint 1. Upload a image and draw black color on area you want to inpaint. For example, we want to fix the hands of this image

2. Select the generative fill style and write the prompt to describe your desired output. 3. Click Draw and That's it. Result:

### Outpaint 1. Upload a black image or upload a random image and draw it to all black. Add image to the canvas

2. Select the generative fill style and write the prompt to describe your desired output. 3. Click Draw and That's it. Result:

# Edit with mask To edit a specific area of an image while keeping the rest untouched, you can use a mask. By drawing over the target area, you define exactly where the AI should apply changes and where it should remain unchanged. When an image contains multiple subjects, such as a group of people, editing just one via text prompts alone can be challenging, as the AI may inadvertently alter others. By using a mask, you provide the AI with precise spatial guidance, ensuring the modifications are restricted only to your selected subject.

The result is as below:

# Z Image

Z-Image is a powerful and highly efficient image generation model with **6B** parameters. **Model Highlights**: * **Photorealistic Quality**: Delivers strong photorealistic image generation while maintaining excellent aesthetic quality * **Accurate Bilingual Text Rendering**: Excels at accurately rendering complex Chinese and English text * **Prompt Enhancing & Reasoning**: Prompt Enhancer empowers the model with reasoning capabilities * **Sub-second Inference**: Achieves fast generation speed on supported hardware # Image to Video Currently, we have three models for Image-to-Video: [**Hunyuan Video**](/video/hunyuan-video/image-to-video), [**Wan2.1 Video**](/video/wan2.1-video), and **Kling Video**. Please upgrade to the latest version to get all three models. Please visit [www.dreamerland.ai](https://docs.dreamerland.ai/video/www.dreamerland.ai) for details. | Model | iOS | Android (Google Play) | Android (APK) | | ------------- | --- | --------------------- | ------------- | | Kling | ✅ | ✅ | ✅ | | Wan2.1 Video | ✅ | ❌ | ✅ | | Hunyaun Video | ✅ | ✅ | ✅ | * **Kling Video** produces the highest-quality output but has restrictions on sensitive content and lacks LoRA support. * [**Wan2.1 Video**](/video/wan2.1-video) is the best alternative to Kling. While its quality is slightly lower, it offers more creative freedom. * [**Hunyuan Video**](/video/hunyuan-video/image-to-video) is the fastest among the three, generating decent video output in just 1–2 minutes. It also supports LoRA styles. # Hunyuan Video We are thrilled to announce the addition of Hunyuan Video to our platform! This cutting-edge text-to-video model represents a significant leap forward in the realm of video generation, and we can't wait for you to explore its incredible capabilities. Please download the latest version of Dreamerland on app store or download the latest android APK. ### What is Hunyuan Video? Hunyuan Video is currently the most parameter-rich and high-performance Text-to-Video and [Image-to-Video](/video/hunyuan-video/image-to-video) model in the open-source domain. This immense capacity enables it to generate videos with exceptional physical accuracy, remarkable scene consistency, and the ability to seamlessly transform static images into dynamic, lifelike videos. Whether starting from text or an image, Hunyuan delivers unparalleled creative possibilities. ### **How it works** For Image to Video feature, please refer to [Hunyaun Image to Video](/video/hunyuan-video/image-to-video) You can find the Hunyuan Video model in the Video tab. We offer various video lengths, ranging from 2 seconds to 4 seconds. Simply input your prompt and optionally select the Lora you wish to apply for styling. The processing time for the video is approximately 1 to 2 minutes, depending on the length you choose.

### Key Elements of a Good Prompt 1. **Start with the Main Subject**: * **Example**: "A talented guitarist strumming a classic acoustic guitar." 2. **Set the Scene**: * **Example**: "On a sunlit beach with gentle waves lapping at the shore." 3. **Detail the Action**: * **Example**: "The guitarist passionately plays a lively tune, fingers dancing over the strings." 4. **Incorporate Camera Instructions**: * **Example**: "The camera slowly pans from the ocean to focus on the guitarist’s expressive face." 5. **Define the Mood and Lighting**: * **Example**: "Golden hour sunlight bathes the scene, casting a warm glow and long shadows in the sand." ### Tips for Crafting Vivid Scenes 1. **Choose Descriptive Language**: * **Weak**: "A busy city" → **Strong**: "A bustling metropolis alive with energy and color." 2. **Ensure Visual Cohesion**: * **Example**: Keep the lighting warm and inviting, with hues of gold and orange reflecting throughout the scene. 3. **Use Clear and Direct Descriptions**: * **Example**: "The camera captures a close-up of the guitarist’s hands, clearly showing the intricate finger movements on the strings." Prompt: A sleek, silver sports car races down a winding mountain road. The backdrop features lush green hills under a clear blue sky. The camera tracks alongside the car, highlighting its speed and elegance. The bright sunlight glints off the car's surface, enhancing the thrill of the moment.

Prompt: A playful puppy chasing after a fluttering butterfly in a lush garden. The setting is filled with blooming flowers and greenery. The puppy bounds excitedly, its ears flopping with each leap. The camera follows closely, capturing the joyful energy of the chase. Sunlight sparkles off the dew on the grass, adding a magical touch to the moment.

### More Examples Below we show some examples generated from Hunyuan Video

# Image to Video We’re excited to introduce the **Hunyuan Image-to-Video** feature in our app! This powerful tool uses AI to turn static images into dynamic videos effortlessly. Perfect for creators and storytellers, it unlocks new ways to bring your visuals to life. Elevate your content with Hunyuan Video today! Currently, the Hunyuan Image-to-Video feature supports **start frame input** as its primary mode of operation. Simply upload your desired start frame image, and let the AI transform it into a video. **Tips:** For optimal results, choose the style that best aligns with the motion you envision. This approach significantly enhances the success rate and ensures greater stability in your video output, delivering smoother and more coherent results. Not all the styles/LoRAs in the app can deliver the desired results, as most of them are specifically trained for text-to-video generation.

### Travel Prompt Seamlessly blend multiple prompts in a single generation! Specify exact start times (e.g., "0:15") to control when each prompt takes effect, creating dynamic, evolving visuals. Perfect for storytelling, scene transitions, or layered artistic effects. Example prompts as below 0s-3s: walking, 3s-5s: waving hand, 5s-7s: adjusting hair, 7s afterward: dancing

{% embed url="" %} result video {% endembed %} # Wan2.1 Video {% embed url="" %} Wan2.1 is a cutting-edge open-source video generation model that outperforms both open-source and commercial solutions in benchmarks. The model excels in Text-to-Video and Image-to-Video. Notably, it’s the first video model with robust Chinese/English visual text generation. Wan2.1 combines high performance, accessibility, and multi-task versatility, advancing open-source video AI.

## **Prompt** Writing prompts for Wan Video differs from writing prompts for Stable Diffusion models. Instead of using a list of tags, you should write in natural language. Focus on describing the actions and motions in the scene to achieve the best results for image-to-video generation. The basic formula for crafting an effective prompt is **Subject + Motion + Scene**. Here's how it works: 1. **Subject**: Clearly define the main focus of your video, such as a person, object, or animal. 2. **Motion**: Describe the action or movement you want to emphasize, as this is crucial for image-to-video generation. 3. **Scene**: Set the context by detailing the environment, background, or setting where the action takes place. For example: * *"A knight (Subject) riding a horse through a forest (Scene) with a flowing cape (Motion)."* * *"A butterfly (Subject) fluttering its wings (Motion) in a sunlit garden (Scene)."* This structure ensures clarity and helps the AI generate more accurate and visually compelling results. ## Frames A start frame is mandatory, while an end frame is optional. Including both a start frame and an end frame can significantly enhance the consistency and coherence of the video. The size of the input image plays a crucial role. If the input image is too small, the quality of the generated video will suffer substantially. High-quality start frames are essential for producing high-quality videos. The better the input, the more refined and visually appealing the output will be. Always ensure your start frame is clear, detailed, and appropriately sized for optimal results. Below is examples of 2 frames and the final output video:

{% embed url="" fullWidth="false" %} Result {% endembed %} ## **Styles** **Tips:** For optimal results, choose the style that best aligns with the motion you envision. This approach significantly enhances the success rate and ensures greater stability in your video output, delivering smoother and more coherent results. Some styles are trained on lower resolutions, which may result in artifacts when used in Professional Mode. If the output appears blurry in Professional Mode, reducing the strength of the style can help improve the clarity and overall quality of the result. Below is the example using Live wallpaper style to animate an image

{% embed url="" %} Output result video {% endembed %} # Pose reference {% embed url="" %} Final result {% endembed %} In image to video mode, you can use a reference video to guide the pose transitions. For best results, make sure the starting pose in your reference video closely aligns with the initial frame of your input image. 1. Upload a start frame 2. Switch on the "Pose reference video" 3. Upload a reference video for the pose

# Video to Video We’ve introduced Video-to-Video in WAN 2.1, allowing you to transform your source videos with AI. Simply upload your input video, enter a text prompt describing your desired output, and let the model generate a new version while preserving motion and composition. 1. Upload a video source 2. Select Control type (Pose, Depth, Face or Canny) 3. Enter the prompt

Here is the result of the example using this sample video and prompt: {% embed url="" %} # Reference Image {% embed url="" %} Prompt: a beautiful girl in kimono. she is in forest. she adjusts hair and bent over. {% endembed %} In Video-to-Video mode, you can enable the Reference Image option to guide the AI's output. When activated, the system will analyze your provided reference image and incorporate its visual style, colors, and composition elements into the generated video while maintaining the original video's motion and timing. The Attention Mask allows you to selectively focus on specific areas of your reference image. By masking a particular region, you direct the AI to prioritize and maintain detail in that area while processing the content.

# Key frames {% embed url="" %} The Start Key Frame allows you to establish precise initial conditions for your video generation. By setting this key frame, you ensure the AI begins processing with your exact desired composition, style, or subject placement.

# Inpaint {% embed url="" %} **Introducing Wan2.1’s Video Inpainting Feature!** Effortlessly swap or remove objects in videos with Wan2.1’s powerful inpainting feature. Perfect for edits, corrections, or creative enhancements, it delivers seamless results. Transform your videos with precision—no complex tools needed. **Steps** 1. Upload a source video

2. Select the object you want to replace or remove. For best results, refine the mask by adding more points to key areas—such as the arms, legs, body, and head—ensuring consistent tracking throughout the entire video.

3. Select Free mode or Control mode. Free mode will ignore the control type. You can see the example above for the different output.

3. Input the prompt to describe the inpainted video 4. (Optional) Add a reference Image. In the above example, we use this reference image.

5. (Optional) Add a key frames. You can add a start frame or end frame to further control the style and structure of the final output video. It is useful especially for removing object. Below is result without using the reference image and start frame. {% embed url="" %} # Free mode By default, video inpaint is restricted by control type so that the inpainted object will maintain the pose or shape of the original video If you want to completely change the structure of the inpainted object, you can use free mode to ignore the control type.

Here is the example of what free mode can achieve: {% embed url="" %} Now, the inpainted object won't be restricted by the pose or shape. We can turn the red car into tiger, horse or woman. # Points and Negative points Video masking differs from image masking because it requires tracking the target object across all frames to ensure accurate inpainting. Rather than manually drawing a mask, you simply specify the object you want to track by adding points, and the system handles the rest. We want to change the Tony's shirt to spider-man suit in the below video. {% embed url="" fullWidth="false" %} We’ll mark positive and negative points to mask his shirt. Positive points indicate the areas you want to select, while negative points define regions to exclude. By placing a few points on his shirt, we can generate an mask like this.

After placing a few points on his shirt, the system masks Tony's entire body. However, since we want to preserve his face, we’ll add negative points to exclude it from the mask. We can get an mask like this.

Now, we enter prompt "a man wears spider man suit" and hit a Start button. {% embed url="" %} # Lip sync # Move Move allows users to specify exactly how objects should move in generated videos by drawing the motion paths. 1. Upload a start frame

2. Draw motion paths. In this example, I want to make her move her head and both hands

3. Write your prompt

And that's it. You can hit the start button and wait for the result. You can click the show motion paths button in the result

{% embed url="" %} # Wan2.2 Video {% embed url="" %} We're thrilled to introduce Wan2.2, the latest generation of our AI model, now available in the app! Experience a new level of creative control with Wan2.2's superior prompt adherence, allowing it to follow your instructions with greater precision. You’ll also notice enhanced visual quality and richer detail, bringing your ideas to life with professional-grade results. Best of all, you can enjoy these powerful upgrades without any extra cost—the Credit and Power usage for Wan2.2 remains exactly the same as the previous 2.1 version. Select Wan2.2 and start creating today!

Please note that certain styles for Wan2.1 might not be compatible with Wan2.2. We have labeled Wan2.2 accordingly. There is a 2.2 on the top left corner for Wan2.2 styles.

# Character Animation The model can animate any character based on a action video, precisely replicating the facial expressions and movements to generate highly realistic character videos. It can also replace characters in a video with animated characters, preserving their expressions and movements while replicating the original lighting and color tone for seamless integration into the environment.

We now offer two animation methods for you, each serving a different purpose. Please check out our [Wan-Animate vs. SCAIL-2 comparison](/video/wan2.2-video/character-animation/wan-animate-vs-scail2) for more information. ## **Wan-Animate** To begin the animation process, upload two files: 1. A Character Image (The subject you wish to animate). 2. An Action Video (The driving motion/pose). Your character will be animated to precisely match the pose and movement from the action video. We use this character in our example:

And the action video: {% embed url="" %} We want to replace the woman in the video with our character. Now, we have uploaded character image and action video. We enabled hands, background and relighting options. for more details, you can visit [this page](/video/wan2.2-video/character-animation/wan-animate-options)

Prompt can be very simple like "a girl is dancing". Below is the final result. {% embed url="" %} # Wan Animate vs SCAIL2 #### ⚡ TL;DR: Which one should you choose? * 🎬 Choose Wan-Animate if: You are animating human and you want exact 1:1 pose transfer. * 🚀 Choose SCAIL-2 if: You are animating either human or non-humans (monsters, animals, robots) or need to handle complex 3D actions like wild spins, flips, and multiple characters interacting without glitches. * If Standard mode, better use SCAIL-2 for low res generation. Face is always degraded in Wan-Animate in low res. #### What is AI Motion Transfer? In simple terms, motion transfer means "borrowing the movement from a video and applying it to a still photo." All you need to do is upload a character image (it could be a real person, an anime character, or a monster) and a "driving video" of someone dancing or moving. The AI will then make your character mimic those exact movements! Right now, the two most powerful and trending technologies for this are **Wan-Animate** and **SCAIL-2**. #### ⚖️ Core Comparison | **Feature** | **🎬 Wan-Animate (The Bone/Skeleton Route)** | **🚀 SCAIL-2 (The Direct AI Route)** | | ---------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | How it Works | It first tracks the human skeleton lines and facial features in the video, then forces your image character to match that skeleton. | No skeleton required. It drops the video and image straight into the AI's brain to match and mimic them pixel-by-pixel. | | Biggest Strength |

Lifelike expressions & perfect swapping

Facial details like glances, blinks, and smiles feel incredibly natural and soulful. Its "Replacement Mode" can swap a character inside a video while perfectly keeping the original lighting and background.

Ultimate freedom & high-difficulty action

Because it isn't trapped by a human skeleton, it easily handles 3D spins, multi-person interactions, and even non-human shapes.

| | Weakness / Limit | When facing complex 3D spins or overlapping limbs, the skeleton tracking can get confused, causing the body to distort or glitch. | While big actions look amazing, tiny, subtle facial expressions (like eye movements) might not feel quite as detailed as Wan-Animate. | **🎯 Choose Wan-Animate when:** 1. Focusing on "Facial Expressions" & Emotion: If your video has lots of close-ups, talking, singing, or needs strong eye contact, Wan-Animate looks the most natural. 2. Seamless Character Swapping (Video-to-Video): If you want to replace the actor in an existing video with your own character image while keeping the exact background and studio lighting. 3. Standard Human Movements: Normal walking, dancing, talking, and standard human poses. **🎯 Choose SCAIL-2 when:** 1. The Character is Non-Human: You want to animate a cat, a monster, a dragon, or a robot. Since they don't have human skeletons, only SCAIL-2 can handle them smoothly. 2. High-Difficulty 3D Actions: The video involves massive spins, flips, rolling on the ground, or limbs crossing over each other. 3. Multiple Characters: There are two or more people interacting, touching, or hugging in the frame. SCAIL-2 handles physical space much better. 4. Fast, Low-Res Drafts: At lower resolutions, SCAIL-2 stays highly stable and doesn't warp or glitch as easily as Wan-Animate. #### Example: Input: {% columns %} {% column %} Character image

{% endcolumn %} {% column %} Pose Video {% file src="/files/FfQZZxNfAUfLZsSiqtxJ" %} {% endcolumn %} {% endcolumns %} #### **Character Replacement** {% embed url="" %} **Pose Transfer** {% embed url="" %} # Wan Animate Options You can customized the video with different options

## Hands Aligning Hands and Fingers from Action Video {% embed url="" %} But sometimes, you don't want your character to copy the finger movements. This usually happens when the character is non-human. In the example below, the cat with human hands and fingers looks unnatural. {% embed url="" %} ## Facial Expression Transfer the facial expression from action video to the character {% embed url="" %} ## Background Use the background of the action video. If this feature is turned off, it will show the background of the character image. {% embed url="" %} ## Relighting Relight the character to match the color tone and lighting of action video. {% embed url="" %} # Anchor Frame An **anchor image** is a reference image used to maintain visual consistency across different segments of the generated video. The model constantly refers back to this anchor to ensure subjects and visual style remain stable. # Seedance 2.0 Director-Level Video Generation Welcome to an all-new video creation experience! Our app now fully supports Seedance 2.0, bringing you a cinematic AI creation tool right at your fingertips. Whether you are starting with a simple idea, a beautiful photo, or need precise control over characters and camera movements, Seedance 2.0 turns your imagination into reality. Let's dive into how to master its three core features: Text to Video, Image to Video, and the incredibly powerful Reference to Video. * [Text to Video](/video/seedance-2.0/text-to-video) - Using text only to create the video * [Image to Video](/video/seedance-2.0/image-to-video) - Using the input image as first frame to create a video * [Reference to Video](/video/seedance-2.0/reference-to-video) - Using input images or video as reference, to create a video with the reference elements # Text to video # Image to video # Reference to video # Minimax H3 MiniMax H3 is MiniMax's latest omni-modal video model, and it's now fully available in Dreamerland. Unlike earlier video models that only read a text prompt or a single image, H3 understands **text, images, video and audio together** — you can hand it a prompt, a few reference photos, a video clip and a music track all at once, and it fuses them into a single coherent shot. Every H3 clip comes with **native audio generated together with the picture** — dialogue, sound effects and ambience are created in the same pass, so there's no separate audio step and no audio toggle to worry about. Clips run from **4 to 15 seconds**, and you can chain them together with the Extend task to build longer scenes. ### Choosing a Task There are 6 Minimax tasks in Dreamerland: * Text to Video * Image to Video * [Reference to Video](/video/minimax-h3/reference-to-video) * Extend Video * [Video Inpaint](/video/minimax-h3/video-inpaint) * [Music Video](/video/minimax-h3/music-video) {% tabs %} {% tab title="Text" %} **Text to Video** The classic starting point — describe the scene and H3 generates it from scratch, sound included. Because H3 generates audio natively, it pays off to describe what you want to *hear* as well as see. {% endtab %} {% tab title="Image" %} **Image to Video** Bring a still image to life. Upload a **start frame**, an **end frame**, or both: * **Start frame only** — the video begins from your image and animates forward. * **End frame only** — H3 generates a shot that lands exactly on your image. * **Both** — H3 creates a smooth transition between the two frames. Great for morphs, reveals and before/after shots. {% endtab %} {% tab title="Reference" %} [**Reference to Video**](/video/minimax-h3/reference-to-video) The most powerful H3 task. Instead of using an image as the literal first frame, you feed H3 **references** and it builds a new shot around them: * **Up to 5 reference images** — lock a character's face, an outfit, a product or an art style. * **A reference video** — borrow motion, choreography or camera work from an existing clip. * **A reference audio** — drive the clip with a voice recording or sound. On H3, an audio reference alone is a valid job: upload a voice line and let the model perform it. {% endtab %} {% tab title="Extend" %} **Extend Video** Continue a video you've already made. Upload the previous clip, choose how the extension should be driven (**text**, **first/last frame**, or **reference**), and H3 picks up where the clip left off — the extension automatically inherits the source clip's aspect ratio. Turn on **Merge source video** if you want the result delivered as one combined clip (original + extension); leave it off to receive just the new part. This is how you build stories longer than 15 seconds: generate a shot, extend it, extend again. {% endtab %} {% tab title="Inpaint" %} [Video Inpaint](/video/minimax-h3/video-inpaint) Edit an existing video with a text instruction — no manual masking needed. Upload the source clip and type a **mask prompt** describing what to change, e.g. *"replace the red car with a white horse"* or *"change her jacket to a leather one"*. * **Control type — None:** the region is freely regenerated from your prompt. * **Control type — Pose:** the original motion is preserved while the appearance changes — perfect for swapping characters or outfits without losing the performance. You can also attach up to 5 reference images to tell H3 exactly what the replacement should look like. The output keeps the source clip's length and aspect ratio. {% endtab %} {% tab title="Music" %} [**Music Video**](/video/minimax-h3/music-video) Purpose-built for making videos cut to a song. Upload a full track (up to 5 minutes) and H3 generates a clip synchronized to it — visuals that move with the music. Because each generation is still one 4–15 second clip, you build the full video piece by piece: 1. Upload your song and generate the first clip (elapsed length = 0). 2. For the next clip, set **Elapsed video length** to how many seconds you've already covered (e.g. 10). 3. Optionally attach the **previous clip** so the new one continues seamlessly from it — same characters, same scene, next part of the song. 4. Repeat until the song is covered, then stitch the clips in the built-in video editor. Reference images and a reference video are optional here too, so your performer can stay consistent across every cut. {% endtab %} {% endtabs %} ### Official Prompt Guide You can find the official prompt guide below: Text to Video and Image to Video: Reference: You can also use our [AI Director](/video/minimax-h3/ai-director-kumi) to help you create prompt following official prompt guide # Reference to Video Reference to Video is MiniMax H3's most powerful task — and the one that makes it a true omni-modal model. Instead of animating an image directly (that's Image to Video), you hand H3 **references** — images, a video clip, an audio track, or any mix of them — and it builds a brand-new shot *around* them. Your character's face, your product, a dance from another video, a recorded voice line: H3 reads them all as one context and fuses them into a single coherent video. {% embed url="" %} \ **Reference to Video vs. Image to Video** The distinction matters, because they solve different problems: * **Image to Video** uses your image as the *literal first frame*. The video starts from exactly that picture. * **Reference to Video** uses your images as *ingredients*. The character from your photo can appear mid-scene, from a new angle, in a new outfit, doing something entirely new — the composition is invented fresh from your prompt, which is also why this task gives you an **aspect ratio** control while Image to Video doesn't. Rule of thumb: if you want *this exact picture to move*, use Image to Video. If you want *this character / thing / style in a new shot*, use Reference to Video. ### The Three Reference Types #### Image references (up to 5) Lock the identity of anything: a character's face, an outfit, a product, a pet, a location, or an overall art style. H3 keeps them consistent throughout the clip. You can combine subjects — for example, two character photos plus one background photo — and direct them in the prompt: *"the woman from \

Use Case	Video Mask Prompt
Replace the entire person	person
Change the face and hair of a person	head, hair
Replace the woman on the right	woman on the right
Replace a red car	red car