Model
Choose a model for the kind of change you need. A fast free model is useful for testing direction; credit models may provide different style control, fidelity, speed, or output options.
Img2Img workflow guide
Make Img2Img Changes You Can Direct
Start with an image, name what must remain recognizable, and describe the exact change. This practical Img2Img guide combines a working generator with prompts, settings, examples, and fixes you can use on the next result.
- Begin with a visual anchor
- Direct one concrete change
- Compare before you publish
Working generator
Try the Img2Img Workflow
Upload a source you own or may use, describe what should stay and what should change, then choose an available model and generation settings. The tool keeps the workflow beside the guide so you can apply each idea immediately.
Generated result
Definition and practical limits
What is Img2Img?
Img2Img is image-to-image generation: an existing image provides the visual starting point, while a prompt and model settings direct what should change.
It is useful for restyling portraits, developing sketches, placing products in new scenes, and exploring variations while keeping selected visual anchors recognizable.
- Accepted source formats
- 3 formats: PNG, JPEG, and WebP
- Maximum upload size
- 10 MB per source image
- Maximum image edge
- 8,192 pixels
Reviewed byImg2Img product team
PublishedUpdated
Product informationAbout Img2ImgContact
Official referencesOpenAI image generation guideGoogle Gemini image generation guide
A repeatable process
How Img2Img Works
Img2Img is most useful when you treat the source as evidence, not decoration. Give the model a clear visual anchor, direct a specific change, and review whether the result preserved the details that matter.
Upload a readable source
Choose a clear photo, product shot, sketch, illustration, or previous AI result. A readable subject and intentional framing give the model a stronger starting point.
Name what stays
State the subject, identity, pose, silhouette, camera angle, layout, logo placement, or other visual anchor that must remain recognizable.
Describe the change
Specify a concrete new setting, material, lighting treatment, palette, medium, or mood. One coherent direction is easier to judge than a list of unrelated ideas.
Compare and revise
Check the output beside the source. Review identity, anatomy, text, logos, proportions, and product details, then revise the smallest instruction that caused the miss.
Control the output
Choose Img2Img Settings With Intent
Settings should support the job you are doing. Start with the source ratio when composition matters, use one output while refining the prompt, and move to higher-quality or additional outputs only after the direction is working.
Available ratios, output counts, quality levels, and credit costs depend on the selected model. The free model shows the full GPT Image 2 settings catalog; only its supported settings remain selectable, while unavailable choices are visible but disabled.
Aspect ratio
Match the source ratio when preserving framing is important. Change the ratio only when the new crop is part of the brief, and expect the model to reinterpret space near the edges.
Output count
Generate one image while fixing the prompt. Multiple outputs are useful after the instruction is stable because they let you compare variations without changing the brief.
Quality and cost
Use low-cost previews to test composition and direction. Increase quality for final detail only after the source, prompt, model, and ratio are producing the right result.
Prompt structure
Write Img2Img Prompts That Are Easy to Judge
A useful Img2Img prompt separates the visual anchor from the requested change. This makes the result easier to review and gives you a clear sentence to revise when the model moves too far or not far enough.
Preserve
Name the subject, identity, pose, silhouette, product shape, room layout, or camera angle that must stay recognizable.
Change
Describe the new environment, lighting, material, wardrobe, palette, medium, or treatment in concrete visual terms.
Style
Add a coherent visual direction such as editorial photography, handcrafted clay, graphic novel ink, or warm architectural rendering.
Check
State constraints worth reviewing: keep facial identity, preserve label placement, retain window positions, or avoid changing the camera angle.
Build a campaign scene
Keep the package shape, label placement, logo, and front camera angle recognizable. Place the product in a bright editorial studio with warm side light, tactile fabric, and a bold seasonal color story.
Use this promptChange the treatment
Keep the face, expression, pose, hairstyle, and crop recognizable. Restyle the portrait as a polished contemporary anime illustration with clean linework, vivid color blocks, and natural skin detail.
Use this promptDevelop an unfinished idea
Preserve the character silhouette, horns, flower, staff, outfit, and front-facing pose. Develop the sketch into a warm, detailed fantasy illustration with soft forest light and handcrafted texture.
Use this promptTroubleshooting
Fix Common Img2Img Results
Do not rewrite everything after a weak result. Identify whether the miss came from the anchor, requested change, framing, or fine details, then adjust that part of the instruction.
The subject no longer looks recognizable
Repeat the visual anchor near the start of the prompt. Name identity, pose, silhouette, or product features precisely, and reduce change strength when the selected model exposes that control.
The result barely changes
Replace broad words such as “better” or “creative” with visible choices for light, material, setting, palette, or medium. Ask for one stronger direction instead of several mild ones.
The composition drifts
Restate camera angle, crop, subject position, and layout. Use an aspect ratio close to the source unless a new composition is intentional.
Hands, text, logos, or details are wrong
Treat these as review points, not guaranteed outputs. Regenerate or revise with the exact detail named, compare at full size, and verify brand assets and text before publishing.
Judge visible change
What a Useful Img2Img Comparison Shows
Strong examples make both preservation and transformation visible. The source should still explain the result, while the new image clearly earns its difference in setting, medium, material, or mood.
Tested in this workspace · September 29, 2026
One source image, one focused style change
We ran this source image through the free Img2Img model with the exact instruction below. Compare the hat, hair, pose, and lace cardigan before looking at the new linework and fabric detail.
- Model
- GPT Image 2 (Free)
- Settings
- 1:1 · 1024×1024 · Low quality · 1 image
Exact prompt used
Keep the woman's short brown hair, wide-brimmed hat, white lace cardigan, pose, framing, and warm background recognizable. Refine this image into a polished hand-painted anime illustration with clean linework, natural facial proportions, soft warm light, and detailed fabric texture. Do not add or remove objects.
In this result, the main subject and framing stay recognizable; the face and clothing details were reinterpreted. Review those details against your own source before using a result.
This is one successful run, not a guarantee that later generations will match it exactly.
Multi-reference portrait fusion into one scene
Three separate portraits from different locations are combined into a single cohesive party photo while keeping each person's facial identity and expression recognizable.
Portrait photo to illustrated style
The face, pose, and framing remain easy to compare while linework, color, texture, and background language move into a new style.
Supported AI Image Models
Switch Between Flagship Image-to-Image Models
Compare Google DeepMind, OpenAI, Alibaba Qwen, ByteDance Seedream, Black Forest Labs Flux, and xAI Grok models in one unified image-to-image workspace.
Compare all AI models
Specialized AI Photo Editing Tools
Direct Workflows for High-Frequency Photo Edits
AI Image Editor with Prompt
Edit photos by text prompt—change backgrounds, outfits, and lighting.
Photo to Sketch AI
Convert portraits and architecture into pencil sketches, ink line art, or manga.
Old Photo Restoration AI
Repair scratches, tears, glare, and faded colors in vintage family photos.
Remove Shadow from Photo AI
Erase harsh facial shadows, phone cast shadows, and uneven lighting.
Remove Filter from Photo AI
Neutralize heavy color tints and restore natural skin tones and true colors.
AI Watermark & Text Remover
Cleanly remove proof watermarks, date stamps, and overlaid text.
AI Expand Image (Outpainting)
Uncrop and extend vertical photos into 16:9 widescreen banners seamlessly.
Img2Img questions
Answers Before You Generate
The practical details users need when choosing a source, writing a prompt, and reviewing an Img2Img result.
What does Img2Img mean, and is it the same as photo to photo AI or pic to pic AI?
Img2Img is short for image-to-image generation—also searched as photo to photo AI, pic to pic AI, upload image AI generator, or AI image from image. An existing image provides the visual starting point, while a prompt and model settings direct a new result. The source can guide subject identity, pose, shape, composition, color, or other visual evidence.
How is Img2Img different from text-to-image?
Text-to-image starts from words alone. Img2Img starts from both an image and an instruction, so it is better suited to restyling, concept variations, sketch development, product scenes, and edits where some visual structure should remain recognizable.
What should an Img2Img prompt include?
Name what should stay, describe the concrete change, add one coherent style or treatment, and list the details you will review. Avoid vague requests and conflicting styles. A focused prompt is easier to compare against the source and easier to revise.
Will Img2Img preserve the original exactly?
No generative model guarantees an exact copy. A clear source and precise anchor improve continuity, but you should still review identity, anatomy, text, logos, proportions, and product details. Use deterministic editing tools when exact pixel-level preservation is required.
Can I try the Img2Img generator for free without signing in?
Try GPT Image 2 (Free) once daily without signing up. Use credits for more images. Other models and advanced settings require VIP access.
Can I use an Img2Img result commercially?
Commercial use depends on your rights to the source, the selected model terms, and the final content. Use images you own or may use, review brand assets and third-party rights, and confirm the applicable model and product terms before publishing.
Start with the source
Put the Img2Img Workflow Into Practice
Upload one clear image, write one focused change, and compare the result against the visual anchor you chose to preserve.