H0R12on

Video Creator/Motion designer

[H0R12on @SungHsu_]

@ragxxy Thanks for sharing so much. Looking forward to your next project!

[H0R12on @SungHsu_]

@ragxxy Great work!
So how do you keep anatomy and motion so accurate in GPT-Image-2 and Seedance 2?
I often get extra/missing fingers, distorted limbs, or unnatural movement. Any prompt tips or workflow tricks?

[H0R12on @SungHsu_]

@ragxxy That sounds fantastic, really looking forward to your next project.
I love this growth concept and have tried it myself. AI often struggles with big size differences and keeps breaking proportions.
How do you handle that?
Thanks for your time, and sorry for all the questions!

[H0R12on @SungHsu_]

@ragxxy Sounds subtle. Have you ever used any skin texture prompts to improve the rendering of foot skin details?

[H0R12on @SungHsu_]

@ragxxy Looks so pro! Let's talk about your content. As we know, The gpt-image-2 model and other i2i or i2v always ban NSFW prompt,such as Hiyuki’s foot scene, it may trigger gpt‘s limit, so how did you avoid it?

[H0R12on @SungHsu_]

@ragxxy So how do you control the influence of the first and last frame images during video generation, so they don’t interfere with motion performance?

[H0R12on @SungHsu_]

@ragxxy NICE! Actually, In my past project, the model often keeps the initial pose or the ending pose for a short moment before the intended animation actually starts or after it ends, which makes the motion feel delayed or “stuck.”

[H0R12on @SungHsu_]

Hi friend! I created character sheets for Hiyuki and Aleph-1 using GPT-Image 2 and always included them as references in every video generation. For a majority of the generations I didn’t generate any key frames. I just wrote the prompt and specified to extend the video starting from the last frame. I did generate key frames using GPT-Image 2 for lots of scenes after Hiyuki kicks Aleph-1 away and it pans down her body from her head to her feet (Wide low angle shot tilted upward behind her body from the head to her back). I generated key frames too for when she takes off her sandals, (the starting frame with her sandals on, and the end frame with her being barefoot.) The other scenes that required me to generate multiple key frames was when she was growing and her foot was flattening him (First frame Aleph-1 holding up her foot, last frame her foot bigger and Aleph-1 lowering to the ground) (Rinse and repeat two more times to get the whole 3 growth spurt sequences). The last scene where she lifts her foot up and shows the flattened one also needed key frames generated for the first and last frames.

[H0R12on @SungHsu_]

@ragxxy Amazing! Did you use GPT-Image-2 (or another image model) to create the storyboard/keyframes first, and then use Seedance 2 for the video generation?If you don't mind sharing, I'd love to learn a bit about how you structured the prompts and the overall pipeline.

[H0R12on @SungHsu_]

@guanyuhan426 这种　覆面警察　执法在法律上是否合法？

[H0R12on @SungHsu_]

私は日本語を食べます

[H0R12on @SungHsu_]

https://t.co/gd5senefsG

[H0R12on @SungHsu_]

扫码查看终末地配队方案

[H0R12on @SungHsu_]

《临行事项》预抽卡活动 https://t.co/4cgnNC6ity

[H0R12on @SungHsu_]

终末地有加好友的吗
エンドフィールドを遊んでるの方はいませんか？

[H0R12on @SungHsu_]

新年明けましておめでとう！

[H0R12on @SungHsu_]

雨の中で、大阪の夜桜を見ます

[H0R12on @SungHsu_]

奈良特別の道路表示

[H0R12on @SungHsu_]

尾道の聖地巡礼

[H0R12on @SungHsu_]

いつか、しまなみ海道の全てをクリアするまで