Specifications
- Input
- text, images
- Output
- images
- Resolution
- 1024:1024, 1080:1080, 1168:880, 1360:768, 1440:1080, 1080:1440, 1808:768, 1920:1080, 1080:1920, 2112:912, 1280:720, 720:1280, 720:720, 960:720, 720:960, 1680:720
How it works Gen4 Image?
Frequently asked questions about Gen4 Image
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
gen4-image is a Runway model for image generation. It works in two modes: creating an image from a text description and processing an uploaded image, where the reference sets the character, scene, or style. It is suitable for illustrations, concept art, advertising visuals, and series of frames with consistent characters.
The output is raster images ready for publication. Frame proportions are set in the prompt, so it is easy to adapt one plot for a feed, banner, or print, and the image-to-image mode preserves the composition and style of the reference. See the request settings for exact resolution values.
There is no separate step-by-step reasoning mode in our data, and we will not claim it exists. The model solves a different task: it turns a text description and references into an image, following the details of the request. If you need long chains of reasoning, it is better to use a text model.
This capability is not in our data, so we will not promise it: gen4-image is a visual model, not a tool for function calling or structured JSON output. Its result is an image based on a prompt or reference. For scenarios with function calling in BotHub, text models are suitable; you can switch between them in one window.
Images, yes, this is part of the model's work: the image-to-image mode uses the uploaded frame as a reference. Audio and video are not declared among the supported inputs. It's simple: provide a text description or an image and get an image.
We have no information on how the model works with Russian-language prompts, so we will not praise it for this. It parses the scene description by meaning; if the wording does not yield the desired frame, try rewriting the request or using a reference image. Access to the model from Russia is available without a VPN or foreign card.