Flux 2 Klein 4B - Text to Image
ComfyUI in the Cloud
Deploy Flux 2 Klein 4B - Text to Image, from
$0.60 / hour
Flux.2 Klein 4B Text-to-Image ComfyUI Workflow
A ComfyUI text-to-image generation workflow powered by Flux.2 Klein 4B, featuring Qwen 3 4B text conditioning, Flux.2 VAE decoding, configurable image dimensions, and seed-based image generation.
Workflow Overview
This workflow utilizes the Flux.2 Klein 4B model for AI image generation within ComfyUI. It combines the Flux.2 Klein 4B diffusion model, Qwen 3 4B text conditioning, and the dedicated Flux.2 VAE to generate images from natural-language prompts.
The workflow provides a text prompt input, two numeric dimension controls, and a noise seed parameter. These inputs allow users to configure the generation request while keeping the main model components organized within a reusable ComfyUI subgraph.
The displayed configuration includes two numeric inputs set to 1024, a noise seed of 0, and a dedicated image output connection. Sampling steps, CFG scale, and sampler settings are not exposed in the visible subgraph configuration.
Generation Settings
Dimension Input 1
1024
Dimension Input 2
1024
Noise Seed
0
Model
Klein 4B
Text Encoder
Qwen 3 4B
VAE
Flux.2
Generation Type
Text-to-Image
Output
IMAGE
The two numeric inputs are labeled value and value_1 in the displayed node. Both are configured to 1024 and may represent image width and height, depending on their internal subgraph connections.
Model Configuration
Image Generation Model (UNet)
Flux.2 Klein 4Bflux-2-klein-base-4b
Flux.2 Klein 4B serves as the primary image generation model responsible for the denoising process. The model uses conditioning information derived from the input prompt to guide image generation. Its 4B model configuration provides a compact alternative to larger image generation models, making it suitable for workflows where model size and resource requirements are important.
Text Conditioning Model
Qwen 3 4Bqwen_3_4b
The Qwen 3 4B model provides text conditioning for the Flux.2 Klein image generation process. It processes natural-language instructions and produces representations that help guide the diffusion model. This conditioning stage enables the workflow to interpret descriptive prompts and use them as guidance during image generation.
Variational Autoencoder (VAE)
Flux.2 VAEflux2-vae
The Flux.2 VAE is responsible for converting generated latent representations into visible images. After the diffusion model completes the image generation process, the VAE decodes the resulting latent representation into pixel-space output. The decoded image can then be connected to preview, saving, or additional image processing nodes.
Text Prompt Configuration
The workflow includes a dedicated text prompt field for providing natural-language instructions to the Flux.2 Klein 4B generation model. Users can describe the subject, composition, environment, lighting, and other visual characteristics they want the generated image to contain.
Prompt Input
The prompt field is empty in the displayed configuration. Text entered into this field can be processed by the Qwen 3 4B conditioning model and used to guide the image generation process.
The prompt input allows the same workflow structure to be reused for different generation requests without changing the selected diffusion model, text encoder, or VAE.
Noise Seed Configuration
The noise seed controls the random initialization used during image generation. The displayed workflow uses a noise seed value of 0.
| Noise Seed | 0 |
| Seed Control | Configurable |
| Randomization Control | Available |
Using a fixed seed can help reproduce results when the remaining model configuration and generation parameters are unchanged. Changing or randomizing the seed produces a different random initialization and can lead to different generated images.
Additional Parameters
| Workflow Type | Text-to-Image |
| Dimension Input 1 | 1024 |
| Dimension Input 2 | 1024 |
| Noise Seed | 0 |
| Prompt Input | Empty |
| Output Type | IMAGE |
| Sampling Configuration | Not exposed in the visible subgraph |
Surrounding Nodes and Workflow Integration
The Flux.2 Klein 4B text-to-image subgraph consolidates model selection, text conditioning, image dimension controls, and noise seed configuration into a reusable image generation component.
The workflow begins with a text prompt describing the desired image. The Qwen 3 4B conditioning model processes the prompt and provides conditioning information for the Flux.2 Klein 4B diffusion model.
The diffusion model uses this conditioning information during the image generation process. The numeric dimension controls can be used to configure the generated image size, depending on their internal connections. The noise seed controls random initialization.
After generation, the Flux.2 VAE decodes the latent representation into a visible image. The generated image is exposed through the IMAGE output connection on the subgraph.
The IMAGE output can be connected to downstream nodes for image preview, saving, or additional processing. Exact internal sampling nodes and downstream connections are not visible in the screenshot.
Workflow Highlights
- ✓ Flux.2 Klein 4B: Uses the Flux.2 Klein 4B diffusion model for text-to-image generation.
- ✓ Qwen 3 4B Conditioning: Uses Qwen 3 4B to process text instructions and provide conditioning for image generation.
- ✓ Flux.2 VAE: Converts generated latent representations into visible images.
- ✓ Configurable Dimensions: Includes two numeric inputs set to 1024, suitable for image dimension configuration when connected accordingly.
- ✓ Seed-Based Generation: Supports configurable noise initialization with the displayed seed set to 0.
- ✓ Text Prompt Input: Provides a dedicated field for natural-language image instructions.
- ✓ IMAGE Output: Exposes generated images for preview, saving, or further processing.
- ✓ Reusable Subgraph: Consolidates model configuration, prompt input, dimension controls, and image output into a single component.
Workflow Summary
This ComfyUI workflow combines the Flux.2 Klein 4B diffusion model with Qwen 3 4B text conditioning and the Flux.2 VAE for text-to-image generation. The configuration includes a dedicated prompt field, two numeric inputs set to 1024, and a noise seed value of 0.
The Qwen 3 4B model processes text instructions, while Flux.2 Klein 4B performs image generation. The Flux.2 VAE decodes the resulting latent representation into an image exposed through the IMAGE output connection.
The subgraph organizes the primary model components and generation inputs into a reusable workflow. Additional sampling settings and internal node connections depend on the complete ComfyUI graph.