CivArchive
    Amphibia - Style and Characters LORA - v1.0
    NSFW
    Preview 141095896
    Preview 141095877
    Preview 141095895
    Preview 141095864
    Preview 141095890
    Preview 141096039
    Preview 141095891
    Preview 141095888
    Preview 141095886
    Preview 141095867
    Preview 141095848
    Preview 141095849
    Preview 141095852
    Preview 141095859
    Preview 141095875
    Preview 141095873
    Preview 141095878
    Preview 141095879
    Preview 141095869
    Preview 141095876

    I will improve this model in the future.

    I spent a week creating this LoRA. This is my first attempt to locally train an ANIMA version. The model has two versions:

    • large for good result for characters and style

    • small for mostly style because characters looks bad in this version.

    The dataset for this LoRA includes 29 characters, group interactions, and backgrounds from the animated series.

    Characters:

    1. Anne Boonchuy, Anne Boonchuy superform, Anne Boonchuy birdform

    2. Sasha Waybright, Sasha Waybright superform

    3. Sprig Plantar

    4. General Yunan

    5. Hop Pop Plantar

    6. Polly Plantar

    7. Marcy Wu, Marcy Wu superform

    8. Valeriana

    9. Leopold Loggle

    10. Albus Duckweed

    11. The Core

    12. Lysil and Angwin

    13. Maddie Flour

    14. Alastair

    15. Bella

    16. Bessie

    17. Captain Grime

    18. Curator Ponds

    19. Felicia Sundew

    20. King Andrias

    21. Mayor Toadstool

    22. Mire

    23. Mother Olm

    24. Mr. Boonchuy

    25. Mrs. Boonchuy

    26. Parisia

    27. Renee Frodgers

    28. Sadie Croaker

    29. Soggy Joe

    They are not reproduced perfectly, but it is what it is. The style can also be applied to completely different characters if you want.

    Description

    Comments (8)

    YogDagenAug 28, 2026· 2 reactions
    CivitAI

    Let's go amphibia

    KklodyAug 28, 2026
    CivitAI

    I'm having a problem with Sprig Plantar and Hop Pop Plantar, and with the creature characters in general. When I write the trigger for the world only, or change their clothes, it messes up or they transform into something different. Even when I tried the Prompt I used in the Sprig Plantar image, it worked fine, but when I changed the clothes, it broke. Or when I removed "goggles on hat, green vest, green shorts," the character changed and became legless. Is it possible to make it work when I write only the character's trigger, like Anne Boonchuy and Sasha Waybright? They work really well when I write only their trigger, even without adding all the details, which gives more flexibility in changing clothes and such with other characters besides Anne and Sasha.

    Because this Lora is really great, and the style is fantastic, I really hope it works without problems. It looks like your work is great, so don't expect me to criticize; I just hope it improves. Keep it up!

    Foxntail
    Author
    Aug 28, 2026· 2 reactions

    Hmm, interesting. Maybe it’s because the Civitai generator uses BF16 rather than the FP32 model. The FP32 model is larger and performs significantly better.

    To be honest, I haven’t experimented with the Sprig and Hop clothing elements, so I’ll check that out. I might re-upload the LoRA with the larger version.

    I’m not at home at the moment, but I’ll definitely test it when I get back. Thanks! I’m glad you liked it overall!

    KklodyAug 28, 2026

    @Foxntail I made sure I'm using FP32 on my computer.

    So keep trying it out and let me know if you have any tips on using the other characters. If so, please tell me how to use the non-human characters; maybe I'm typing the prompt wrong or something.

    Foxntail
    Author
    Aug 28, 2026

    @Kklody I tried changing his clothes to a tuxedo and it worked. I strongly recommend using a full character description in the prompt (sprig plantar, 1pink frog. sprig plantar, pink skin, large eyes, closed mouth. some clothes) and splitting the prompt into blocks:
    amphibia style, character (examples in images). emotions. clothes. pose. background.

    In the dataset I captioned 90% of the images by hand, separating everything with periods. I still don’t fully understand how models work during training in general. When there are that many characters in the dataset, the chance of errors is high. Many characters have very few images where they appear alone, so I had to compensate by increasing the number of repeats for each character. To give you an idea, here is how many references there were per character:
    Backgrounds - 93

    Anne Boonchuy - 56

    Group shots - 54

    Sasha Waybright - 17

    Sprig Plantar - 8

    General Yunan - 7

    Marcy Wu - 7

    Hop Pop Plantar - 6

    Polly Plantar - 6

    Anne Boonchuy superform - 6

    Valeriana - 6

    Leopold Loggle - 4

    Albus Duckweed - 4

    The core - 3

    Lysil and Angwin - 3

    Maddie Flour - 3

    Sasha Waybright superform - 3

    Mother Olm -2

    Mrs. Boonchuy - 2

    Mr. Boonchuy - 2

    Toadie - 2

    Wally Ribbiton -2

    Parisia - 2

    Renee Frodgers - 2

    Felicia Sundew - 2

    Maggie -2

    Bella - 2

    Curator Ponds - 2

    Marcy Wu superform - 2

    Bessie - 1

    Tritonio Espada - 1

    Alastair -1

    Mayor Toadstool -1

    Anne Boonchuy birdform -1

    Sylvia Sundew - 1

    Captain Grime - 1

    King Andrias - 1

    Mire - 1

    Stumpy Stumpson - 1

    Soggy Joe - 1

    Sadie Croaker -1

    ChaosticXAug 29, 2026
    CivitAI

    Thank you!

    FurnomnomSep 4, 2026
    CivitAI

    many thanks for this model !

    Foxntail
    Author
    Sep 4, 2026

    You're welcome!

    LORA
    Anima

    Details

    Downloads
    89
    Platform
    CivitAI
    Platform Status
    Available
    Created
    8/28/2026
    Updated
    9/12/2026
    Deleted
    -
    Trigger Words:
    amphibia style

    Files

    model.safetensors

    Mirrors

    CivitAI (1 mirrors)

    amphibia.safetensors

    Mirrors

    CivitAI (1 mirrors)