<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/cbd26e48-c803-4615-a63d-a57fded2b999/width=525/cbd26e48-c803-4615-a63d-a57fded2b999.jpeg" />VERSION LINKS: <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=936565">FP8 </a>• <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=936309">FP16</a> • <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=942694">NF4 </a>• GGUF <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=936976">Q8_0 </a>/ <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=944704">Q6_K </a>/ <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=944753">Q5_KM </a>/ <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=944736">Q5_KS </a>/ <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=944799">Q5_0 </a>/ <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=944790">Q5_1 </a>/ <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=944898">Q4_KM </a>/ <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=944805">Q4_KS </a>/ <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=944948">Q4_0 </a>/ <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=944903">Q4_1 </a>/ <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=944957">Q3_KM </a>/ <a target="_blank" rel="ugc" href="https://civarchive.com/models/630820?modelVersionId=944952">Q3_KS</a>V2 is an alternative to V1 with sharper good quality images from 4 steps. This version merges Schnell + Finetuned Dev + Hyper using the same but refined formula of variable block ratios from V1. Check comparison images below!<h3 id="make-sure-to-rename-your-files-after-download-civitai-gives-them-wrong-names!-ubcra4otx"><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fa421fa5-de92-4bc7-bd8b-0d0929fd299c/width=525/fa421fa5-de92-4bc7-bd8b-0d0929fd299c.jpeg" />MAKE SURE TO RENAME YOUR FILES AFTER DOWNLOAD, CIVITAI GIVES THEM WRONG NAMES!</h3><h3 id="tested-samplerscheduler-for-low-steps:-a92w8r2r1">Tested sampler/scheduler for low steps:</h3><ul><li>ComfyUI: euler sampler, simple or beta scheduler.</li><li>Forge: euler, flux realistic sampler. KL Optimal or beta scheduler.</li></ul>This model doesn't take guidance parameter, like schnell.<h3 id="the-versions-with-aio-(all-in-one)-in-the-name-include-unet-+-vae-+-clip-l-+-t5xxl-(fp8).-also-known-as-checkpoint-or-compact-version.-nb2sqzld3">The versions with AIO (All in one) in the name include UNET + VAE + CLIP L + T5XXL (fp8). Also known as Checkpoint or Compact version.</h3><h3 id="using-bnb-nf4-and-gguf-quants-in-comfyui-requires-installing-custom-nodes-that-add-special-model-loaders:-jpm5c9rdg">Using BNB NF4 &amp; GGUF quants in ComfyUI requires installing custom nodes that add special model loaders:</h3><ul><li>GGUF: <a target="_blank" rel="ugc" href="https://github.com/city96/ComfyUI-GGUF">https://github.com/city96/ComfyUI-GGUF</a></li><li>NF4 + Lora support: <a target="_blank" rel="ugc" href="https://github.com/bananasss00/ComfyUI_bitsandbytes_NF4-Lora">https://github.com/bananasss00/ComfyUI_bitsandbytes_NF4-Lora</a></li></ul><ul><li>(outdated) NF4 UNET: <a target="_blank" rel="ugc" href="https://github.com/DenkingOfficial/ComfyUI_UNet_bitsandbytes_NF4">https://github.com/DenkingOfficial/ComfyUI_UNet_bitsandbytes_NF4</a></li><li>(outdated) NF4 AIO checkpoint: <a target="_blank" rel="ugc" href="https://github.com/comfyanonymous/ComfyUI_bitsandbytes_NF4">https://github.com/comfyanonymous/ComfyUI_bitsandbytes_NF4</a></li></ul><h3 id="for-using-unet-versions-you-also-need-to-have-the-text-encoders-and-vae.-8xmo5hs9y">For using UNET versions, you also need to have the TEXT ENCODERS and VAE.</h3>If you don't have them, download them from here:<ul><li>T5XXL - CLIP L: <a target="_blank" rel="ugc" href="https://huggingface.co/comfyanonymous/flux_text_encoders/tree/main">https://huggingface.co/comfyanonymous/flux_text_encoders/tree/main</a></li><li>GGUF T5XXL: <a target="_blank" rel="ugc" href="https://huggingface.co/city96/t5-v1_1-xxl-encoder-gguf">https://huggingface.co/city96/t5-v1_1-xxl-encoder-gguf</a></li><li>VAE: <a target="_blank" rel="ugc" href="https://huggingface.co/black-forest-labs/FLUX.1-schnell/tree/main/vae">https://huggingface.co/black-forest-labs/FLUX.1-schnell/tree/main/vae</a></li></ul>Place the model in "models/diffusion_models" or "models/unet", both text encoders in "models/clip" and vae in "models/vae" folder.In ComfyUI, use the standard flux workflow or add 'Load Diffusion Model', 'DualClipLoader' and 'Load VAE' nodes to replace the checkpoint loader and complete the setup.In Forge, set the option "Diffusion in low bits" to "bnb-nf4"<pre><code>Thanks to city96 for gguf quantization script. 
Thanks to reddit user a_beautiful_rhind for bnb quantization script.</code></pre><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5e4ab077-7e0e-43ef-b9c6-5b7d5db13849/width=525/5e4ab077-7e0e-43ef-b9c6-5b7d5db13849.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/411ee655-605e-40c1-a2df-8e04c28a2b89/width=525/411ee655-605e-40c1-a2df-8e04c28a2b89.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/6eab35e8-c5e5-4247-8b25-73c6175c70cd/width=525/6eab35e8-c5e5-4247-8b25-73c6175c70cd.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/6882c984-f9f5-4119-a201-74d10d8ffd96/width=525/6882c984-f9f5-4119-a201-74d10d8ffd96.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/130893ca-3016-4a7d-aa48-286cfd60fd7a/width=525/130893ca-3016-4a7d-aa48-286cfd60fd7a.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/855faf80-025e-487d-a978-75ed2d828e5c/width=525/855faf80-025e-487d-a978-75ed2d828e5c.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5b6907ee-d0a9-42bd-92db-a95b6da97567/width=525/5b6907ee-d0a9-42bd-92db-a95b6da97567.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/8efda954-0908-4464-961e-bd077573ecca/width=525/8efda954-0908-4464-961e-bd077573ecca.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/4dd836dc-15e5-4676-849f-fc1e342e5b34/width=525/4dd836dc-15e5-4676-849f-fc1e342e5b34.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/bb621b2e-2251-4035-a4c0-74d190429c48/width=525/bb621b2e-2251-4035-a4c0-74d190429c48.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/37b3cbdc-d093-40b2-92c1-1f48446d68f5/width=525/37b3cbdc-d093-40b2-92c1-1f48446d68f5.jpeg" /><hr /><h1 id="flux-fusion-version-1-wouhi2tw2">FLUX FUSION VERSION 1</h1>Merge of Schnell and Dev variants of the Flux.1 model with a irregular smoothed ratio for each of the layers.Quick comparison between versions. Prompts and settings at the end.↓↓ Click show more for more examples and instructions ↓↓<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/c5f66651-0a1e-4f15-9c1e-6f306452796d/width=525/c5f66651-0a1e-4f15-9c1e-6f306452796d.jpeg" />Recommended use around 8 steps. If textures like skin look overworked, try lowering steps.<h3 id="comparison-of-v1-quants:-l6x1fgqr1">Comparison of V1 QUANTS:</h3><h3 id="test-parameters:-8-steps-cfg-3.5-1536x1536-seed-0-shum1zymn"><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/54a5c150-d169-4394-9208-a76dd92bc702/width=525/54a5c150-d169-4394-9208-a76dd92bc702.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9572a0ac-8896-4c8f-a7a5-a61685e9ddec/width=525/9572a0ac-8896-4c8f-a7a5-a61685e9ddec.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/a0766b14-4a8f-49a3-9a0c-d41cf44b2cf0/width=525/a0766b14-4a8f-49a3-9a0c-d41cf44b2cf0.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fd4e856e-88af-41da-acec-8cb7010216e4/width=525/fd4e856e-88af-41da-acec-8cb7010216e4.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/8a2a98f1-387f-45b0-a0df-0a8d7230231d/width=525/8a2a98f1-387f-45b0-a0df-0a8d7230231d.jpeg" />Test parameters: 8 Steps, CFG 3.5, 1536x1536, seed 0</h3><h3 id="prompts:-vehergd7l">Prompts:</h3><ol><li>"Extreme closeup, frog face, star crystal structure, intricate designs, glowing hues. Extreme depth of field, celestial light, shimmering details, otherworldly charm, majestic elegance."</li><li>"extremely detailed 3d render portraits of a cyber-dragon themed flaming gothic arcane tech woman, cables, arcane tech-dragon inspired design, exposed machinery. the casing is glittery transparent tinted orange, red and black, allowing to see the internals. sophisticated fantasy design. abstract thematic background. extreme depth of field. dragon behind"</li><li>"classic pokemon 3DCG illustration. thick outlines. pokemon render style. extremely dynamic composition showcasing the special ability power. flowing pose with extreme closeup on the face in the upper area of the image. intense perspective, in motion movement effects, dynamic impactful vfx, eye catching.. A rock Pokémon with earthquake powers poses dynamically, its body a twisted mass of rugged terrain and molten lava flows. The upper area zooms in on its face, a mask of stone and fury with blazing eyes. Earth shatters beneath its feet as it stomps, unleashing seismic waves that ripple through the abstract background like a fractured canvas. Intense perspective compresses space, conveying unstoppable power. Vibrant colors dance: fiery oranges, electric blues, and smoldering grays. Movement effects blur edges, blurring boundaries between rock and energy. Impactful VFX burst forth in the foreground, echoing the Pokémon's raw force.. masterpiece, professional, best quality, sharp, extreme detail, Hyper-detailed, high-resolution, intricate, vivid. "</li><li>"Ethereal female face in 4K ultra closeup, eyes radiating eerie mystical aura with crystalline composition tinted purple-blue hues. Surrounding inferno blazes with dynamic flames and motion effects, creating a vertiginous atmosphere. Extreme depth of field emphasizes surreal otherworldly presence. Glowing eyes at the focal point contribute to haunting mystique, shot from an altered viewing angle emphasizing mysticism. Use Octane and Redshift raytracing for realistic fire and light effects, achieving ultra-realistic 3D render with intense, dreamlike quality."</li><li>"Ethereal star princess, diaphanous gown, shimmering stardust, intricate halo, luminous beauty. Night sky, glowing constellations, soft light, dreamy ambiance, mesmerizing allure."</li><li>"Sci-fi landscape, derelict alien structure, holographic iridescence, massive metal arches, dark skies, damaged antennas. Ground littered with debris, scattered wreckage, distant moon, dim light."</li></ol>

Flux Fusion V2 [4 steps] [GGUF •  NF4 • FP8/FP16]

Merge between Schnell and Dev variants of the Flux.1 model, with an irregular curve of ratios between blocks.Recommended 4 to 8 steps from testing.VAE and CLIP included in .safetensors file, fp8.

Files with SHA256: f280536ac99b7b096d82529dec5cff1c25c99b57f85581a04e43194f20a5d2e7