CivArchive
    Drift stop from close-up face (WAN2.2 I2V) (ANIME) AKIRA in Kaneda's Bike Drift Stop | AKIRAのバイクスライド - Hight_WAN2.2_I2V
    NSFW

    ⚠️This LORA is for WAN2.2 I2V video generation!!⚠️

    The sample PNG image is meta information for video generation. Drag and drop it into ComfyUI to view the workflow and resources.



    This is the meme action from "Akira" where Kaneda's large motorcycle skids to a halt.

    The starting image should be a close-up of a face, and it will probably work with any vehicle you specify in the prompt.

    It is intended to be able to generate video from any close-up face image.

      

    1. FPS:24 & Num Frames:49 is Best Setting.

    2. Please start with a close-up of her face as the starting image.

    3. Please specify the vehicle she will be riding at the beginning of the prompt.

    4. If you can set the end image, you can create the perfect video you imagine❤️

    Specifying an end image is not required. If you specify only a start image, enjoy the whims of the random WAN2.2 seed gods🙏



    Landscape 1216x832 .... ❤️Highly recommended❤️

    Portait 832x1216 ............ ⚠️Not recommended.⚠️

    Square 1024x1024 ......... ⚠️Not recommended.⚠️

      

    training dataset is here.

    https://mega.nz/folder/t4JBkTxK#vYORIzus-3CnCsAZIDcubw



    Basic Motorcycle Prompt

    She's riding a sideways Motorcycle.
    Super fast animation.
    She is always looking at the camera.
    She's moving at super speed.
    She glides off into the background, sending up sparks and trailing dust.
    She and her vehicle come to a stop sideways.

     

    Any Vehicle Prompt

    She's riding a sideways  <***Any vehicle***>.
    Super fast animation.
    She is always looking at the camera.
    She's moving at super speed.
    She glides off into the background, sending up sparks and trailing dust.
    She and her <***Any vehicle***> come to a stop sideways.

    Replace <***Any vehicle***> with any vehicle.



    sorry Japanese text

    顔面のクローズアップから、乗り物に乗って画面奥にスライドして停止する。

    ただ、それだけです。

      

    トレーニングしたクリップはほぼバイクですが、顔のアップから始まるため、どんな乗り物にでも(ある程度は)対応できます。

     

    1. FPS:24 Num Frames 49 の2秒動画を作成するのが最適だと思います。

    2. 開始画像は必ず顔のアップシーンから開始してください。

    3. 乗り物は、プロンプトの先頭行で指定してください。

    4. 終了画像を指定できるなら、おそらくイメージしたままの動画が生成されるはずです。

      

    たぶん…おそらく...WAN2.5や最近話題の中国系動画生成サービスを使えば、プロンプトだけで同様のシーンは生成できるかも(?)しれないですね。

     

    本モデルの作成目的は、顔面のクローズアップ画像という「汎用的な素材から任意のアクションを生成する」という、汎用モデルの…まぁええか。

     

    このモデルは、トレーニングの最適化手順テストでつくった健全モデルです。

    このモデルで解説記事を書く予定でなので、参考資料としてアップロードしました。

     

    今後とも、よろしくお願いします。



    モデル作成には、AIモデル作成支援ソフト『HENTAI APP』を利用しています。

    詳細は以下記事を参照:https://civarchive.com/articles/26629/create-wan22-datasets-from-videos-optimized-software-has-been-released


    📘rticle from the WAN2.2 IV2 Training Guide

    https://civarchive.com/articles/27705



    私は本LORAの利用に、制限・権利・責任を一切主張しません。

    Description

    FAQ

    Comments (8)

    phasod139Mar 4, 2026· 8 reactions
    CivitAI
    suteakasuteakasuteka434
    Author
    Mar 4, 2026

    Thank you!!!!!!!!!!!!!!!!!!!!!!!!!🥹

    LyaraMar 5, 2026· 2 reactions
    CivitAI

    A really great and dynamic Lora-Idea! I wonder how it would work out if you would have a closeup with someone wearing a shield, blocking an attack and sliding back on her feet. It would make a great crossover example of what's possible with this! : ))

    suteakasuteakasuteka434
    Author
    Mar 6, 2026· 2 reactions

    @Lyara

    Thank you for your comment.

    I've updated the LORA cover illustration to include your idea.

    There are two ways to enjoy LORA:

    - Specify only the start image and enjoy the random generation of WAN2.2.

    - Specify the start and end images to create a perfect video.

    LyaraMar 7, 2026

    @suteakasuteakasuteka434 Greetings and thanks for giving it a try, Sir! Aaand you did a great job on that again as well: it is turning out exactly like I imagined it to hopefully work out! It's like it was straight out of an anime already 💜
    And yes, for the most of the scenes I had in mind I will be having a before and after pic. But I'm gonna render it in 3D, as there's more control over the consistency this way, so WAN only has to do the movement and not a lot of guesswork in-between. Yet I'll be having some "open end" rendering approaches too, because it's fun to see where the Lora might take your picture, if you give it some freedom too. *smirks*

    Anyyys: I am so looking forward to work with this for my action-oriented scenes; 😇thank you again for being as kind as sharing it online! 🫶🫶 I'll be sharing an example as soon as I got the renderings done all in all. : )

    nodoguroMar 7, 2026· 1 reaction
    CivitAI

    アニメの歴史に残るダイナミックなシーンをloraにしていただいて、ありがとうございます!

    いろいろな乗り物で試してみたいと思います。

    suteakasuteakasuteka434
    Author
    Mar 8, 2026· 1 reaction

    コメントありがとう!WAN2.2のトレーニングはまだまだ勉強中です。新しい実用的で正解が曖昧な技術の試行錯誤って本当に楽しいです。

    …まぁ、実用目的の「実用」は実に卑猥で猥褻なエロ目的なのですが…エロ目的の勉強は本当に楽しいです❤

    SamurMetro97Mar 9, 2026· 1 reaction
    CivitAI

    楽しいLoRAに感謝です!

    LORA
    Wan Video 2.2 I2V-A14B

    Details

    Downloads
    237
    Platform
    CivitAI
    Platform Status
    Available
    Created
    3/4/2026
    Updated
    6/18/2026
    Deleted
    -
    Trigger Words:
    She's riding a sideways motorcycle. Super fast animation. She is always looking at the camera. She's moving at super speed. She glides off into the background, sending up sparks and trailing dust. She and her vehicle come to a stop sideways.
    She's riding a sideways <***Any vehicle***>. Super fast animation. She is always looking at the camera. She's moving at super speed. She glides off into the background, sending up sparks and trailing dust. She and her vehicle come to a stop sideways.

    Files

    AkiraBikeSlide_H_Wan2-2_i2v_A14B.safetensors

    Available On (1 platform)

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