Edit and transform reference images
AI Image to Image Generator
Upload one or more reference images and describe what should change. Use the source for subject, composition, product, or style context, then direct the new environment, lighting, color, materials, or visual treatment with a prompt.
SkyVid supports up to 16 reference images in the current Image to Image workflow. Choose an available ratio and resolution, review the result, and make a more specific follow-up edit when needed.
Transform images with supported AI image models
Image to Image only shows workflows configured to accept reference images. The available resolution tiers, ratios, and credits follow the selected model.
Seedream 5 Pro1k / 2kSeedream 5 LiteSeedream 5 Lite basic / Seedream 5 Lite high
Nano Banana 2 LiteNano Banana 2 Lite
GPT Image 21k / 2k / 4k
Wan 2.7 ImageWan 2.7 ImageNano Banana 21k / 2k / 4k
Grok ImagineGrok ImagineHow to use the AI image to image generator
- 01
Add reference images
Upload between one and 16 images, giving each reference a clear purpose such as subject, product, composition, or style.
- 02
Describe what to keep and change
State the details that must remain consistent, then name the background, light, color, material, or style changes.
- 03
Choose output settings
Select a supported ratio and resolution, generate the result, and make the next edit more specific if needed.
Image to image editing prompt templates
The most useful editing prompts separate invariants from changes. Use these structures to make the source image's role explicit.
Prompt template
Product
Product scene variation
Preserve the product design while replacing the surface, environment, and lighting direction.
Prompt direction
Keep the bottle shape, label layout, cap, and proportions unchanged. Place it on dark volcanic stone beside a shallow reflective pool, cool dawn light from the left, premium outdoor campaign photography, no extra text.
Prompt template
Portrait
Background and atmosphere
State which identity and pose details must remain before describing the new environment.
Prompt direction
Keep the person's face, hairstyle, pose, and clothing unchanged. Replace the background with a quiet train platform at blue hour, soft overhead light, light rain in the distance, realistic documentary color.
Prompt template
Art direction
Style direction
Preserve composition and subject placement while changing medium, texture, and color language.
Prompt direction
Keep the original composition and all object positions. Reinterpret the scene as a cut-paper editorial illustration with visible fiber texture, flat shadows, and a limited navy, lime, and coral palette.
Prompt template
Multiple references
Multi-reference composition
Assign a single job to each image so the references do not compete with one another.
Prompt direction
Use image 1 for the person's face and hairstyle, image 2 for the jacket design, and image 3 for the studio lighting and background colors. Create a waist-up campaign portrait with natural proportions and no logos.
What can you change with image to image AI?
Product scene variations
Keep product details while testing surfaces, environments, lighting, and campaign moods.
Background and atmosphere
Move a subject into a new location, season, time of day, or color environment.
Style and material direction
Reinterpret an existing composition through a new medium, texture, palette, or rendering approach.
Multi-reference concepts
Combine clearly assigned subject, product, wardrobe, composition, and style references in one brief.
How to write an image editing prompt
Begin with what must not change. Then describe the requested edits in visual terms and remove any reference that does not have a clear role.
Keep + change + reference roles + avoid
Define the invariants
Name identity, product details, pose, composition, text, or proportions that must remain stable.
Assign each reference a role
Explain whether an image supplies the subject, product, wardrobe, layout, material, or style direction.
Avoid conflicting instructions
More references are not automatically better. Remove images or directions that compete for the same visual decision.
Image to image upload and output settings
The current workflow accepts multiple reference images and filters ratio and resolution combinations using the selected model's rules.
- Reference images
- Upload 1 to 16 PNG, JPG, WebP, or GIF images.
- File size
- Each reference image can be up to 10MB.
- Prompt length
- Up to 20,000 characters for editing instructions.
- GPT Image 2 limits
- Auto is limited to 1K, and square 1:1 output is unavailable at 4K.
Learn more
Explore practical prompt guides, then return to the generator with a clearer brief.
Image to Image FAQ
What is image to image generation?
Image to image generation creates a new result from one or more reference images plus a written edit prompt. The references can guide subject, composition, product, or style details.
How many reference images can I upload?
The current SkyVid Image to Image workflow accepts between one and 16 reference images in a generation.
What should I tell the AI to keep unchanged?
Name the important invariants directly, such as a face, product shape, label layout, pose, composition, clothing, or color that should remain consistent.
Which file formats and sizes are supported?
SkyVid accepts PNG, JPG or JPEG, WebP, and GIF reference images. Each file can be up to 10MB.
Which aspect ratios and resolutions can I use?
The available settings depend on the selected model. For GPT Image 2 workflows, Auto is limited to 1K and 1:1 output is unavailable at 4K.
How do I get more consistent results from multiple references?
Give each reference one clear role and state what must remain unchanged. Remove images that provide conflicting versions of the same subject, product, or style.