CivArchive
    FLUX.1-Turbo-Alpha - FLUX.1-Turbo-Alpha
    NSFW
    Preview 35849380
    Preview 35842466
    Preview 35843386
    Preview 35843381
    Preview 35843378
    Preview 35843385
    Preview 35845281
    Preview 35845279
    Preview 35845321
    Preview 35845324

    This checkpoint is a 8-step distilled Lora, trained based on FLUX.1-dev model.

    It use a multi-head discriminator to improve the distill quality. the model can be used for T2I, inpainting controlnet and other FLUX related models. The recommended guidance_scale=3.5 and lora_scale=1, Euler = Simple.

    Key Features:

    • Rapid Image Generation: FLUX.1-Turbo-Alpha allows users to create images quickly, making it ideal for projects requiring fast turnaround times

    • Versatility in Applications: This model supports text-to-image generation and inpainting, providing users with the flexibility to create varied and intricate scenes easily.

    • Optimized Settings: For optimal performance, it is recommended to use a guidance scale of 3.5 and a Lora scale of 1, which allows for effective control over the output quality and style.

    • Future-Proofing: Upcoming versions of FLUX.1-Turbo-Alpha will include configurations that utilize fewer steps, further enhancing speed without sacrificing quality.

    Ideal Use Cases:

    FLUX.1-Turbo-Alpha is perfect for artists, designers, and content creators who need quick and high-quality image outputs. Its adaptability to various creative applications makes it a valuable addition to any toolkit, whether for professional projects or personal exploration.

    With FLUX.1-Turbo-Alpha, you can harness the power of AI to elevate your creative projects, transforming ideas into stunning visuals in mere moments.


    Thanks to the original creator :
    alimama-creative
    Original HuggingFace Repo :
    https://huggingface.co/alimama-creative/FLUX.1-Turbo-Alpha

    Description

    Nothing to add, for now.

    FAQ

    Comments (21)

    EKKIVOK
    Author
    Oct 22, 2024· 5 reactions
    CivitAI

    Incredible release up coming guys :)

    trust me, it's incredible <3 you will like it :)

    sevenof9247Oct 22, 2024· 3 reactions
    CivitAI

    it is not an 8step turbo

    on hugging-faceis nothinkg like that discribed ...

    a Lora that makes a picture a little better, whether it's 8 or 45 steps

    EKKIVOK
    Author
    Oct 22, 2024

    Hi thanks for your comment !

    i made several test with that lora with Euler simple 8 Steps VS Euler Simple 28 steps.
    And indeed the quality is less that the regular dev version at 28 steps (higher will be always better).

    BUT, it can do barely the same things in a less time with more precision and qquality (that's normal),

    It can be useful for small configuations that cannot handle a full dev model.

    But you're right, this Lora is in Alpha version, like it was one day for SDXL ^^

    Give it some time :) i hope and i think it will improve again and again :)

    Oh and the description come directly from Hugging Face btw ^^.

    blobby99Nov 1, 2024· 1 reaction

    @EKKIVOK Quote "It can be useful for small configuations that cannot handle a full dev model."

    It's a LORA! It ADDS to a model, not replaces it. Therefore it will be more likely to cause OOM issues for people with smaller amounts of VRAM. Indeed, with ComfyUI, with its darn awful memory-management, one can see so many peeps enquire why adding a LORA ruins their render time.

    EKKIVOK
    Author
    Nov 1, 2024

    @blobby99 it's half true. But here i'm talking about gpu. models and loras with forge are added to RAM.

    there is many configuration with a lot of ram but poor gpu. that system management was not implemented in comfyui. but in Forge, yes. it's called the shared memory, GPU/RAM. so you are in the true but it's 50/50.

    It's for that many people have different result with this lora. cause not everybody have the same computer configuration and settings. It's for that i use "It can be" and not "it will" ^^

    SuzanneOct 23, 2024· 5 reactions
    CivitAI

    👍 Very good lora, it works perfectly and the results are impeccable.

    EKKIVOK
    Author
    Oct 23, 2024· 1 reaction

    Thanks :)

    MoudaHDec 26, 2024· 5 reactions
    CivitAI

    amazing detailed lora, but it's kinda not that consistent when used with some character loras, i found that "ByteDance Hyper-FLUX Acceleration LoRA" works better with character loras than the turbo's alpha sadly, even tho i like the quality of the turbo one's better!

    great work btw buddy i would hope in the upcoming updates it these problems get fixed!! 👏


    EKKIVOK
    Author
    Dec 27, 2024

    the model you mention is the second version of that one, so technicaly it's normal that is a little bit more advanced ^^

    iopsychFeb 20, 2025· 4 reactions
    CivitAI

    anyone have this running with a flux fill model on comfyui? Always get ERROR lora diffusion_model.img_in.weight shape '[3072, 384]' is invalid for input of size 196608

    7145279Feb 23, 2025

    yeah me too, the same, idk why

    othongdipalakpakpol591Mar 3, 2025

    same happened to me (btw my model is "flux1-fill-dev-fp8.safetensors")

    lcsfilmshk169Apr 17, 2025

    same

    alrnr2232Apr 22, 2025

    Any solution or replacement for this?

    EKKIVOK
    Author
    May 1, 2025

    hey ! i think that it came from the model OR the input images used....

    drdewme618May 21, 2025

    did anybody figure this out, i'm getting the same error

    irucangi11636Jun 13, 2025

    check if the checkpoint and the lora are trained on the same model, es FLUX 1

    DiabolicXApr 27, 2025· 3 reactions
    CivitAI

    First of all, I have to thank you for the fantastic job you did, thank you!Speed ​​and excellent image quality are the characteristics of this model. This is a great model especially for those with a weaker GPU and they want to use Flux dev.Thank you once again for this model, which I can recommend to everyone.

    EKKIVOK
    Author
    May 1, 2025· 1 reaction

    <3 thanks

    LoveUrselfJan 13, 2026
    CivitAI

    will try it with my 3060 6g😋

    EKKIVOK
    Author
    Jan 13, 2026· 1 reaction

    It will work :)

    LORA
    Flux.1 D
    by EKKIVOK
    Download (Beta) View on CivitAI
    flux.1
    flux1.d
    turbolora
    photorealistic
    hyper
    base model

    Details

    Downloads
    28,831
    Platform
    CivitAI
    Platform Status
    Available
    Created
    10/21/2024
    Updated
    5/15/2026
    Deleted
    -

    Files

    FLUX.1-Turbo-Alpha.safetensors

    Size:
    661.93 MB
    SHA256:
    77f7523a5e9c3da6cfc730c6b07461129fa52997ea06168e9ed5312228aa0bff

    Mirrors

    HuggingFace (62 mirrors)
    CivitAI (3 mirrors)
    CivitasBay (1 mirrors)
    ModelScope (1 mirrors)
    ModelScope CN (1 mirrors)
    TensorArt (3 mirrors)

    Available On (2 platforms)

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

    SeaArt
    FLUX.1-Turbo-Alpha - FLUX.1-Turbo-Alpha
    CivitasBay
    FLUX.1-Turbo-Alpha - FLUX.1-Turbo-Alpha