What you have to supply
Capability tells you what a model produces. This tells you what it wants first, which is the part that decides whether you can use it today. 15 require a starting image and 12 accept reference images on top; anything not on that list runs from a text prompt alone.
| Model | You must supply | It will also take |
|---|---|---|
| Flux 2 Klein 4B Base Edit | A starting image | Up to 4 reference images |
| Pruna Image Edit | A starting image | Up to 5 reference images |
| GPT Image 1 Edit | A starting image | Up to 6 reference images |
| GPT Image 1.5 | A text prompt only | Up to 16 reference images |
| Qwen Image Edit 2511 | A starting image | Up to 6 reference images |
| Seedream Edit | A starting image | Up to 10 reference images |
| FireRed Image Edit | A starting image | Up to 3 reference images |
| Kling Image O1 | A starting image | Up to 10 reference images |
| Qwen Image 2 Edit | A starting image | Up to 3 reference images |
| GPT Image 2.5 Flare Edit | A starting image | — |
| GPT Image 2.5 Sunburst Edit | A starting image | — |
| Mai Image 2.5 Edit | A starting image | — |
| Nano Banana 2 | A text prompt only | Up to 14 reference images |
| Runway Gen4 Image | A starting image | — |
| Seedream 4.5 | A text prompt only | Up to 10 reference images |
| Topaz Image Colorization | A starting image | — |
| Topaz Upscale Image | A starting image | — |
| HunyuanImage 3.0 Instruct Edit | A starting image | Up to 3 reference images |
Read straight off each model's own record — requires_image, requires_video_input, accepts_video_input, max_reference and the per-mode reference configuration. Nothing is inferred from a model's name or description.
All 18 models
Cheapest complete job first. Every row links to that model's spec page — tier and mode variants share a page with the model they are a variant of, so this is products rather than SKUs.
| Model | Provider | Credits per job | Capabilities |
|---|---|---|---|
| Flux 2 Klein 4B Base Edit | Flux | 1 credit | also edit image |
| Pruna Image Edit | PrunaAI | 1 credit | also edit image |
| GPT Image 1 Edit | OpenAI | 2 credits | also edit image |
| GPT Image 1.5 | OpenAI | 2 credits | also text-to-image and edit image |
| Qwen Image Edit 2511 | Qwen | 2 credits | also edit image |
| Seedream Edit | ByteDance | 2 credits | also edit image |
| FireRed Image Edit | PrunaAI | 3 credits | also edit image |
| Kling Image O1 | Kling | 3 credits | also edit image |
| Qwen Image 2 Edit | Qwen | 3 credits | also edit image |
| GPT Image 2.5 Flare Edit | OpenAI | 4-7 credits | also edit image |
| GPT Image 2.5 Sunburst Edit | OpenAI | 4-7 credits | also edit image |
| Mai Image 2.5 Edit | Microsoft | 4 credits | also edit image |
| Nano Banana 2 | 4 credits | also text-to-image and edit image | |
| Runway Gen4 Image | Runway | 4-7 credits | this only |
| Seedream 4.5 | ByteDance | 4 credits | also text-to-image and edit image |
| Topaz Image Colorization | Topaz Labs | 7 credits | also edit image |
| Topaz Upscale Image | Topaz Labs | 7 credits | also edit image |
| HunyuanImage 3.0 Instruct Edit | Hunyuan | 8 credits | also edit image |
How these models are billed
Five different billing shapes across the roster, which is why a straight comparison of headline figures between two of these models can mislead.
- Charged a flat rate per generation8 models
- Charged per megapixel of output5 models
- Charged per image2 models
- Charged quality resolution flat2 models
- Charged a flat rate that changes with output resolution1 model
All figures are Versely credits. A credit figure marked as a headline rate is not the cost of a complete job. Your plan's credit allowance is on the pricing page.
More capability pages
Every model on Versely · Model providers · Ranked buyer guides
Frequently asked questions
How many image-to-image models does Versely have?+
18 with a spec page — all of them image models, from 11 providers. Tier and mode variants of the same model share a page with the model they are a variant of, so this is products rather than SKUs.
What do image-to-image models need as input?+
15 of the 18 require a starting image, 0 require a video file, 0 accept a video without requiring one and 12 take reference images (up to 16 on the most flexible). Anything not on that list runs from a text prompt alone, and every figure here is read off the model's own record rather than inferred from its name.
What does an image-to-image job cost?+
Complete jobs run 1 to 8 credits across the 18.
Run every one of them on one subscription
The all-in-one AI studio for creators. 60+ models for video, image, voice, music and lipsync — in your browser or on your phone.
Free account. Works in your browser - no install needed. The same account signs in on your phone.