CivArchive
    Preview 24864360
    Preview 24868255
    Preview 24868253
    Preview 24868256
    Preview 24868254
    Preview 24868250
    Preview 24990336
    Preview 24990460
    Preview 24990545
    Preview 24990601
    Preview 24990731
    Preview 24990843
    Preview 24990904
    Preview 24990903
    Preview 24990902

    VERSION LINKS: FP8 FP16NF4 GGUF Q8_0 / Q6_K / Q5_KM / Q5_KS / Q5_0 / Q5_1 / Q4_KM / Q4_KS / Q4_0 / Q4_1 / Q3_KM / Q3_KS

    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!

    MAKE SURE TO RENAME YOUR FILES AFTER DOWNLOAD, CIVITAI GIVES THEM WRONG NAMES!

    Tested sampler/scheduler for low steps:

    • ComfyUI: euler sampler, simple or beta scheduler.

    • Forge: euler, flux realistic sampler. KL Optimal or beta scheduler.

    This model doesn't take guidance parameter, like schnell.

    The versions with AIO (All in one) in the name include UNET + VAE + CLIP L + T5XXL (fp8). Also known as Checkpoint or Compact version.

    Using BNB NF4 & GGUF quants in ComfyUI requires installing custom nodes that add special model loaders:

    For using UNET versions, you also need to have the TEXT ENCODERS and VAE.

    If you don't have them, download them from here:

    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"

    Thanks to city96 for gguf quantization script. 
    Thanks to reddit user a_beautiful_rhind for bnb quantization script.


    FLUX FUSION VERSION 1

    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 ↓↓

    Recommended use around 8 steps. If textures like skin look overworked, try lowering steps.

    Comparison of V1 QUANTS:

    Test parameters: 8 Steps, CFG 3.5, 1536x1536, seed 0

    Prompts:

    1. "Extreme closeup, frog face, star crystal structure, intricate designs, glowing hues. Extreme depth of field, celestial light, shimmering details, otherworldly charm, majestic elegance."

    2. "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"

    3. "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. "

    4. "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."

    5. "Ethereal star princess, diaphanous gown, shimmering stardust, intricate halo, luminous beauty. Night sky, glowing constellations, soft light, dreamy ambiance, mesmerizing allure."

    6. "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."

    Description

    FAQ

    Comments (23)

    akshaydixit007Aug 18, 2024· 2 reactions
    CivitAI

    GGUF is best one :) ... really appreciate your work .. more buzz to you :)

    Anibaaal
    Author
    Aug 18, 2024· 1 reaction

    Thank you!!

    expert78Aug 19, 2024
    CivitAI

    Will we get gguf quants unet only?

    GrannikAug 19, 2024· 1 reaction

    Are there any missing? We have the Q4, Q5, etc.

    Actually, now that I check, new K quants are available! @Anibaaal They (should be) faster and higher quality.

    antonioccostajr881Aug 19, 2024
    CivitAI

    Can I use these models with diffusers libraries on python, to automate image production using code lines?

    antonioccostajr881Aug 19, 2024

    i tried using diffusers library and I get this error:
    OSError: runwayml/stable-diffusion-v1-5 does not appear to have a file named config.json.

    Anibaaal
    Author
    Aug 19, 2024

    @antonioccostajr881 Hi, I'm personally not knowledgeable about diffusers so I'm not sure I can help with that.
    Which merge version are you trying? It could be that the UNET versions have a non-standard naming of the tensors because of the way they were exported, thanks to user city96 for finding out.
    So far only the GGUF files were corrected but the others are in process of reupload, should be updated by tomorrow. It doesn't affect usage in ComfyUI/Forge, but the model might not be recognized properly by other software.

    antonioccostajr881Aug 19, 2024

    @Anibaaal I see... I'm using NF4 all in one version. I'm using the official documentation on this site (https://huggingface.co/docs/diffusers/main/en/api/pipelines/flux) and I didn't find a way to load GGUF models, so I'll wait for the safetensors file with the correct naming. Thanks for updating!

    Anibaaal
    Author
    Aug 19, 2024

    @antonioccostajr881 ohh :/ are you sure NF4 is supported ? you could try fp8/fp16 AIO.
    The all in one versions were correct, only the unet models were non-standard but they have been reuploaded with the correction.

    TarterboxAug 19, 2024· 1 reaction
    CivitAI

    Thanks so much for the quants.

    daniilomaia332908Aug 19, 2024
    CivitAI

    This model v0_bnb-nf4 (AIO) 10.54GB (https://civitai.com/models/630820?modelVersionId=735521), have all versions included? (NF4, All GGUF, FP8 and 16 )

    Is this real?

    If it's true that it has all these all in one it's really incredible

    I had already downloaded model v0-fp8-e4m3fn (AIO) 15.9GB, (https://civitai.com/models/630820?modelVersionId=705611) and found it perfect

    I use the Forge, and work fine !!!

    Congratulations on your work, I'm going to test your model right now

    Anibaaal
    Author
    Aug 20, 2024

    Thank you!

    It doesn't include all quants in one, sorry for the misunderstanding :D

    I think I know what happened, the download file had all those versions in the name? It's because civit ai uses the title from my post for the downloads file name. I edited the title to make it more clear.

    daniilomaia332908Aug 20, 2024· 1 reaction

    @Anibaaal I have used this v0_bnb-nf4 (AIO) 10.54GB in my Forge, to better images I use 6 steps. Work finest. Great job mate. Congratulations

    4489695Aug 20, 2024· 4 reactions
    CivitAI

    Q4 works well for me. Comparable to the base model, faster and more efficient than the base model. The biggest takeaway is good generations even with only 6 steps. The sacrifice is slightly worse results in my experience, but not worse enough to make it not worth it.

    I find it is MUCH faster when I stick to using a cfg of 1.0. Anything different such as 0.8 or 1.2 makes it take twice as long per iteration. Just thought I'd share, found that interesting.

    daniilomaia332908Aug 20, 2024· 2 reactions
    CivitAI

    I have used your model v0_bnb-nf4 (AIO) 10.54GB on my Forge with Diffusion in low bits" to "bnb-nf4. The results are amazing. Which other of your models do you think is best for me to test here?

    BNB NF4/FP4 & GGUF Q4/Q5/Q8?

    I have latest Forge in a RTX 3050 8GB with 32GB RAM

    What's your tip?

    Anibaaal
    Author
    Aug 21, 2024· 1 reaction

    NF4 should be the best for you afaik. GGUF is good too for low vram, but it's a bit slower because it is compressed and the decompression part adds time. Thanks for the tip and trying the merge :)

    AIStudio80Aug 22, 2024

    Hi. I have rtx3050 8gb. My observations are that Although NF4 models are fast, but GGUF models give better results I am using flux1-dev-Q8_0.gguf along with t5xxl text encoder, clip and vae, Processing time is 5 minutes. I am impressed.

    daniilomaia332908Aug 24, 2024· 1 reaction

    @AIStudio80 Wow... 5 minutes and you think it's fast? Here my images are generated in 24s and I still think it's slow :)

    I use the bnb-nf4 (AI) version, I will test this GGUF version that you mentioned and come back here later to comment on the results and image generation times.

    EggbenaAug 21, 2024
    CivitAI

    holy shit... this is confusing, which one is the fastest haha

    Anibaaal
    Author
    Aug 21, 2024· 5 reactions

    Yes it's hard to keep up. Stick to fp8/fp16 if you have enough vram. gguf or nf4 quants are new and experimental and you should use them if you can't fit the others in your GPU, gguf are actually slower than regular fp8/fp16 versions and nf4 is about the same speed but doesn't support lora at least in comfy, I think it does in forge.

    EggbenaAug 21, 2024

    @Anibaaal  yea thanks, I tried a few of the super low qt ones, really the only saved me file size, the 16gb to 10gb models seemed to work best

    sooner or later the cream will rise to the top lol

    very exciting stuff though

    daniilomaia332908Aug 23, 2024· 2 reactions

    @Anibaaal I use your bnb-nf4 (AIO) version with good results, and I support lora in Forge with Diffusion in Low Bits Automatic (fp16 Lora). It works perfectly. Congratulations on the amazing work!

    cathylevermanAug 23, 2024
    CivitAI

    awesome. I wonder if its possible to finetune it so it becomes even better at text?

    Checkpoint
    Flux.1 D

    Details

    Downloads
    221
    Platform
    CivitAI
    Platform Status
    Available
    Created
    8/18/2024
    Updated
    6/12/2026
    Deleted
    -

    Files

    fluxFusionDSNF4GGUFQ4Q5Q8Fp8Fp16_v0GGUFQ5UNET.zip

    Available On (1 platform)

    Same model published on other platforms. May have additional downloads or version variants.