Generate images
These models generate images from text prompts. Here, you can find the latest state-of-the-art image models that fit your use case.
Our picks
Best overall image generation model: google/nano-banana-pro
The best overall image generation model is google/nano-banana-pro. It offers state-of-the-art performance in prompt following, visual quality, text adherence, image detail, and output diversity. This image generation model also has logic built in, allowing you to answer questions with image outputs.
Best fast image generation model: prunaai/p-image
P-image from Pruna AI offers sub-second image generation for less than a cent while still maintaining quality. Additionally, prunaai/z-image-turbo speedily produces stunning image outputs.
Best model for generating images with text: google/nano-banana-pro
Text rendering with google/nano-banana-pro is astonishing and has the ability to handle paragraphs of text.
As a cheaper (and underrated) option, Ideogram models are strong in many areas, but they're especially known for their ability to generate realistic, legible text. See our blog on Ideogram v3.
bytedance/seedream-4.5 is also comparable to nano-banana and is worth giving a shot if you are looking to try a new model.
Best model for generating images with SVGs: recraft-ai/recraft-v3-svg
The Recraft V3 SVG model is the first major text-to-image model with the ability to generate high quality SVG images including logotypes, and icons. The model supports a wide list of styles.
Best ComfyUI model: fofr/any-comfyui-workflow
If you’re a fan of ComfyUI, you can export any of your favorite ComfyUI workflows to JSON and run them on Replicate using the fofr/any-comfyui-workflow model. For more information, check out our detailed guide to using ComfyUI.
Best fine-tunes
Make sure to check out our FLUX fine-tunes collection, which includes all publicly available FLUX fine-tunes hosted on Replicate. This collection should help you get a feel for the sorts of things you can do with fine-tuning.
Featured models
google/nano-banana-proGoogle's state of the art image generation and editing model 🍌🍌
Updated 1 day, 4 hours ago
10M runs
prunaai/z-image-turboZ-Image Turbo is a super fast text-to-image model of 6B parameters developed by Tongyi-MAI.
Updated 2 days, 6 hours ago
12.3M runs
qwen/qwen-imageAn image generation foundation model in the Qwen series that achieves significant advances in complex text rendering.
Updated 2 days, 10 hours ago
1.4M runs
prunaai/p-imageA sub 1 second text-to-image model built for production use cases.
Updated 1 week, 1 day ago
1.9M runs
google/imagen-4-fastUse this fast version of Imagen 4 when speed and cost are more important than quality
Updated 1 month ago
3.8M runs
black-forest-labs/flux-2-maxThe highest fidelity image model from Black Forest Labs
Updated 1 month ago
228.8K runs
bytedance/seedream-4.5Seedream 4.5: Upgraded Bytedance image model with stronger spatial understanding and world knowledge
Updated 1 month, 2 weeks ago
2.1M runs
google/nano-bananaGoogle's latest image editing model in Gemini 2.5
Updated 1 month, 4 weeks ago
74.3M runs
bytedance/seedream-4Unified text-to-image generation and precise single-sentence editing at up to 4K resolution
Updated 2 months ago
23.6M runs
black-forest-labs/flux-proState-of-the-art image generation with top of the line prompt following, visual quality, image detail and output diversity.
Updated 2 months, 1 week ago
13.8M runs
ideogram-ai/ideogram-v3-turboTurbo is the fastest and cheapest Ideogram v3. v3 creates images with stunning realism, creative designs, and consistent styles
Updated 2 months, 2 weeks ago
6.6M runs
black-forest-labs/flux-schnellThe fastest image generation model tailored for local development and personal use
Updated 6 months, 4 weeks ago
599.6M runs
Recommended Models
Frequently asked questions
What’s the fastest model for generating images?
Speed depends on the model’s architecture and how it’s optimized for the hardware it runs on. If you want quick results, models like google/imagen-4-fast and black-forest-labs/flux-schnell are built to return outputs fast, which is great for rapid iteration.
Which model gives the best balance of cost and quality?
Smaller or “fast” variants usually cost less to run. bytedance/seedream-4.5 and ideogram-ai/ideogram-v3-turbo are good picks if you want solid image quality without spending a lot.
What’s the difference between text-to-image and image-to-image?
Text-to-image models create a new image from scratch based on your text prompt. Image-to-image models take an existing image and use your prompt to change or build on it. Think of it as “paint something new” vs. “edit what’s already there.”
Which model makes the most realistic images?
bytedance/seedream-4.5 and ideogram-ai/ideogram-v3-turbo are great for realistic lighting, textures, and faces. They’re popular for lifelike portraits, product shots, and scenery.
Which model is best for artistic or stylistic work?
If you’re aiming for a specific look, black-forest-labs/flux-1.1-pro and black-forest-labs/flux-schnell give you more control over style, lighting, and composition. They’re good for illustrations, concept art, or anything with a creative twist.
Can I edit images with a text prompt?
Yes. Use text-guided editing models like black-forest-labs/flux-kontext-pro or bytedance/seedream-4.5 to add or change details in an existing image. For example, you can tell it to “add sunglasses” or “turn it into a painting.”
What resolution do these models support?
Most models output images between 512×512 and 4K. Check the model card for the exact dimensions supported. Higher resolutions can cost more and take a bit longer to run.
How do I make consistent characters or scenes?
Use a reference image or a fixed seed. Models like black-forest-labs/flux-kontext-pro and ideogram-ai/ideogram-v3-turbo support both, so you can keep the same look across multiple runs.
Can I fine-tune a model with my own data?
Some models support fine-tuning. Look for the fine-tune tag on the model page or check the README for training details.
Can I host my own model on Replicate?
Yes. Push a model from GitHub with a replicate.yaml file. Once it’s built, it runs on the same infrastructure as other models.
Can I use these models for commercial work?
Check the “License” section on the model page. Some licenses allow commercial use, others don’t. Always make sure before using outputs in anything public or commercial.
