CivArchive
    face poser - concept-face_poser-v3-64dim
    NSFW
    Preview 382817
    Preview 382816
    Preview 382815
    Preview 382814
    Preview 382813
    Preview 382881
    Preview 382911

    New version. Basically the same data set, but new training options. I'm not sure it's better.

    Implied fellatio is now a tag it seems to recognize, for more circular shape to the open mouth.

    Share your results! The frames for the new gif are in the training data download.

    this model is meant to help pose faces for lewd scenes

    • open mouth: combine with tongue and uvula for wide open

    • closed mouth: combine with cum in mouth, for dribbling cum

    • teeth: combined with lower and upper to determine mouth pose

    • tongue: overtrained on tongue out, so put tongue out in the negative prompt if you want the tongue in the mouth

    • lower teeth: affects the angle you see the mouth

    • upper teeth: affects the angle you see the mouth

    • closed eyes: eyes weren't a focus, but as you can see in the sample, it helps a lot

    • open eyes

    • rolling eyes

    • one eye closed: wink

    • half-closed eye

    • cum in mouth: combine with cum on tongue in the negative for cum in the back of the mouth

    • oral invitation: makes the pose more direct, can be combined with hand gesture, but the results are small

    • cum on tongue

    • close-up: helps increase detail for inside the mouth, increase strength for inside the mouth,

    • blush? I'm not actually sure what impact this has on blush, it works well, transitions well, but that might just be the AOM3 model

    Let me know what you think about the gif in the preview; do you think they're worth it for showing off models?

    Honestly, it took me almost as long to make that, as it took to train the model.

    Description

    this model is meant to help pose faces for lewd scenes

    • open mouth: combine with tongue and uvula for wide open

    • closed mouth: combine with cum in mouth, for dribbling cum

    • teeth: combined with lower and upper to determine mouth pose

    • tongue: overtrained on tongue out, so put tongue out in the negative prompt if you want the tongue in the mouth

    • lower teeth: affects the angle you see the mouth

    • upper teeth: affects the angle you see the mouth

    • closed eyes: eyes weren't a focus, but as you can see in the sample, it helps a lot

    • open eyes

    • rolling eyes

    • one eye closed: wink

    • half-closed eye

    • cum in mouth: combine with cum on tongue in the negative for cum in the back of the mouth

    • oral invitation: makes the pose more direct, can be combined with hand gesture, but the results are small

    • cum on tongue

    • close-up: helps increase detail for inside the mouth, increase strength for inside the mouth,

    • blush? I'm not actually sure what impact this has on blush, it works well, transitions well, but that might just be the AOM3 model

    Let me know what you think about the gif in the preview; do you think they're worth it for showing off models?

    Honestly, it took me almost as long to make that, as it took to train the model.

    FAQ

    Comments (10)

    emberbyte
    Author
    Apr 2, 2023· 40 reactions
    CivitAI

    If anyone's interested, I can see about writing a script to do prompt list to gif for the webui.

    I don't have one yet, but if there's demand, I can probably knock one out pretty quick.

    Adeptus08Apr 2, 2023

    Would love that! Please reply here if you decide to make it.
    Or even just a suggestion/guide on prompting the gif you made incrementally.

    TroublesomeAApr 2, 2023

    At first I honestly thought it was a model aimed to produce grids for animation, like various bj animation on the site. I'd very much appreciate something like that but with face animation!

    emberbyte
    Author
    Apr 2, 2023· 2 reactions

    @Adeptus08 Sure. Follow me on here, and you should get a notification when I add it. Maybe tomorrow.

    @clex The grid things are pretty cool, but they're super limited because of how low res each frame needs to be.

    If I limit it to a close-up of the face, and train for only four frames at 1024x1024, that would give 512x512 for each frame, but you can't really do much in four frames, or can you?

    If you have any ideas for scenes that would make a good animations with only four frames, then my ears are open. I might make something if I like the idea. And, if I can actually find data for it :)

    sugarcookie69Apr 2, 2023

    looking forward

    1011450Apr 2, 2023

    Please do, it would be unique!

    emberbyte
    Author
    Apr 2, 2023· 1 reaction

    @Adeptus08 I just uploaded it, it's in it's own model upload.

    johnjohnApr 2, 2023

    What about interpolating between frames?

    emberbyte
    Author
    Apr 3, 2023

    @johnjohn you can do that with RIFE; flowframes is a decent free application.

    I might make a script for that too, but that's more difficult, because it'd require the rife model.

    emberbyte
    Author
    Apr 3, 2023

    @johnjohn ah, I misunderstood what you were saying. Interpolating can help for the low frame count, to an extent. The more interpolated frames, the more jank it looks. That's worth considering though. If I take 10 frame long animations, cut them down to 4, it might be enough to rely on interpolation for the final animation.

    LORA
    Other

    Details

    Downloads
    2,198
    Platform
    CivitAI
    Platform Status
    Available
    Created
    4/2/2023
    Updated
    6/11/2026
    Deleted
    -
    Trigger Words:
    wink
    half-closed eye
    open mouth
    closed mouth
    uvula
    teeth
    upper teeth
    lower teeth
    close-up
    rolling eyes