Download
1 variant available
The FLUX.1 [dev] Model is licensed by Black Forest Labs. Inc. under the FLUX.1 [dev] Non-Commercial License. Copyright Black Forest Labs. Inc.
IN NO EVENT SHALL BLACK FOREST LABS, INC. BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.
FLUX.2 Kontext Img2Img — Multi-Reference + Strength-Guidance Image Editing Workflow by TripTrap
Files:
flux-2-klein-9b-kv-fp8.safetensors
qwen_3_8b_fp8mixed.safetensors
Description
This workflow turns FLUX.2's native Kontext architecture into a fully-featured image-to-image
editing powerhouse. Built from the ground up over months of iteration, it gives you surgical
control over how reference images influence your generations — from subtle texture injection to
complete stylistic transformation.
Whether you're editing album art, refining character designs, or exploring glitch aesthetics,
this workflow lets you feed in up to 2 reference images and dial in exactly how much they
influence the output with a single strength slider.
### What makes this different
- FLUX.2 Native — This isn't a hack. It uses FLUX.2 klein 9B KV-FP8's built-in Kontext reference
system for clean, artifact-free image conditioning
- Strength Control — A denoise slider that actually works. 0.0 = no change, 1.0 = full
generation. No cryptic parameters.
- Multi-Reference — Chain up to 6 reference images through the Kontext system for composite influences
- Prompt Engineering — Built-in text concatenation for complex multi-source prompts
- VRAM Safe — Auto-scales images to 0.5MP before encoding, preventing accidental OOM crashes
- 4-Step Generation — Optimized for FLUX.2's distilled inference. Blazing fast.
### How it works
```
Reference Image → Scale → VAE Encode ──┐
├──→ ReferenceLatent Chain → Conditioning → Sampler
Reference Image → Scale → VAE Encode ──┘
↑
Prompt → Text Concat → CLIP Encode → Flux Guidance ──────────┘
```
The reference images are injected directly into FLUX.2's Kontext conditioning system — the same
mechanism the BFL API uses internally. No separate style models, no CLIP vision encoders, no
bloat.
### Requirements
Models (download and place in ComfyUI/models/):
┌────────────────────────────────────┬───────────────────┐
│ File │ Location │
├────────────────────────────────────┼───────────────────┤
│ flux-2-klein-9b-kv-fp8.safetensors │ diffusion_models/ │
├────────────────────────────────────┼───────────────────┤
│ qwen_3_8b_fp8mixed.safetensors │ text_encoders/ │
├────────────────────────────────────┼───────────────────┤
│ flux2-vae.safetensors │ vae/ │
└────────────────────────────────────┴───────────────────┘
Custom Nodes: (install via ComfyUI Manager)
- rgthree-comfy (Power Lora Loader)
- ComfyUI-Custom-Scripts (ShowText)
- ComfyUI_Comfyroll_CustomNodes (CR Aspect Ratio)
### How to use
1. Load the workflow JSON (drag into ComfyUI)
2. Load your base image in Ref 1
3. Optionally load a second reference in Ref 2
4. Write your prompt
5. Set denoise (0.65 is a good starting point)
6. Queue it
### Tips
- 0.5–0.65 denoise: Subtle edits, preserves composition
- 0.65–0.8 denoise: Balanced changes with strong reference influence
- 0.8–1.0 denoise: Heavy transformation, prompt dominates
- Use 4 steps for speed, or up to 8 for more refinement
- Keep images under 1MP for best performance
### License
This workflow is free to use under the FLUX Non-Commercial License.
Respect the model license terms.
────────────────────────────────────────────────────────────────────────────────
▶ YouTube — TripTrap Music
https://www.youtube.com/c/TripTrap1?sub_confirmation=1
▶ Spotify — TripTrap Artist
https://open.spotify.com/artist/7vPfTIL4FweqDqTjcw4Bvr
▶ Apple Music — TripTrap
https://music.apple.com/us/artist/triptrap/1760346233
────────────────────────────────────────────────────────────────────────────────
Built by TripTrap — glitch artist, experimental pop producer, and AI alchemist.
Every image this workflow generates carries a frequency. Make it yours.