FLUX 3 Image Klein distilled text-to-image workflow, slimmed to 9 nodes. This is the fast path: 16 steps, no CFG, no out-painting, just prompt in and native-resolution image out.
Main features:
- Built on the step-distilled FLUX 3 Image Klein checkpoint (flux3_image_klein_sd_int8_convrot), which trades a little fidelity for roughly 3x faster sampling
- 16 steps, no classifier-free guidance, euler/simple, denoise 1.0 - the distilled model is trained to run without classifier-free guidance
- Only 9 nodes: text encode, latent, sampler, decode, save. Nothing to configure beyond the prompt and resolution
- Optional prompt upsampler (qwen3.5_9b_flux3_prompt_upsampler_w4a8) rewrites short prompts into the detailed captions FLUX 3 prefers
- Same latent space and VAE as the base model, so seeds and resolutions are interchangeable between the two graphs
Suggested workflow:
1. Write the prompt. If you keep it short, leave the prompt upsampler enabled - the distilled model punishes vague captions harder than the base does.
2. Set resolution with ResolutionSelector or directly on EmptyFlux2LatentImage (both sides divisible by 16).
3. Run and save. For a quality comparison against the base model, use the same seed and resolution in workflow F07.
Dependencies:
- flux3_image_klein_sd_int8_convrot.safetensors (distilled UNET)
- qwen3.5_9b_flux3_prompt_upsampler_w4a8.safetensors (prompt upsampler)
- qwen3vl_4b_int8_convrot.safetensors (text encoder / CLIP)
- flux2-vae.safetensors (VAE)
- CreateBoundingBoxes, Flux3LayoutPrompt, ResolutionSelector (custom nodes)
RunningHub workflow page: https://www.runninghub.ai/zh-cn/user-center/1846601088051884033/webapp?inviteCode=rh-v1111
Bilibili tutorial video: https://www.bilibili.com/video/BV1btJb6LEGS/
RunningHub registration with invite code rh-v1111: https://www.runninghub.ai/?inviteCode=rh-v1111
Description
FLUX 3 Image Klein distilled text-to-image workflow, slimmed to 9 nodes. This is the fast path: 16 steps, no CFG, no out-painting, just prompt in and native-resolution image out.
Main features:
- Built on the step-distilled FLUX 3 Image Klein checkpoint (flux3_image_klein_sd_int8_convrot), which trades a little fidelity for roughly 3x faster sampling
- 16 steps, no classifier-free guidance, euler/simple, denoise 1.0 - the distilled model is trained to run without classifier-free guidance
- Only 9 nodes: text encode, latent, sampler, decode, save. Nothing to configure beyond the prompt and resolution
- Optional prompt upsampler (qwen3.5_9b_flux3_prompt_upsampler_w4a8) rewrites short prompts into the detailed captions FLUX 3 prefers
- Same latent space and VAE as the base model, so seeds and resolutions are interchangeable between the two graphs
Suggested workflow:
1. Write the prompt. If you keep it short, leave the prompt upsampler enabled - the distilled model punishes vague captions harder than the base does.
2. Set resolution with ResolutionSelector or directly on EmptyFlux2LatentImage (both sides divisible by 16).
3. Run and save. For a quality comparison against the base model, use the same seed and resolution in workflow F07.
Dependencies:
- flux3_image_klein_sd_int8_convrot.safetensors (distilled UNET)
- qwen3.5_9b_flux3_prompt_upsampler_w4a8.safetensors (prompt upsampler)
- qwen3vl_4b_int8_convrot.safetensors (text encoder / CLIP)
- flux2-vae.safetensors (VAE)
- CreateBoundingBoxes, Flux3LayoutPrompt, ResolutionSelector (custom nodes)
RunningHub workflow page: https://www.runninghub.ai/zh-cn/user-center/1846601088051884033/webapp?inviteCode=rh-v1111
Bilibili tutorial video: https://www.bilibili.com/video/BV1btJb6LEGS/
RunningHub registration with invite code rh-v1111: https://www.runninghub.ai/?inviteCode=rh-v1111