@theocox.bsky.social
muscle watcher and maker of AI pics and videos. Find me also on DeviantArt (DA): Deviant Art: https://www.deviantart.com/plubby
24.01.2026 14:33 — 👍 26 🔁 7 💬 0 📌 0
23.01.2026 14:30 — 👍 15 🔁 3 💬 0 📌 0
22.01.2026 19:50 — 👍 38 🔁 6 💬 1 📌 0
18.01.2026 13:18 — 👍 33 🔁 5 💬 1 📌 0
17.01.2026 21:37 — 👍 41 🔁 7 💬 0 📌 1
17.01.2026 01:37 — 👍 23 🔁 4 💬 0 📌 0
16.01.2026 17:26 — 👍 21 🔁 3 💬 1 📌 0
07.01.2026 23:46 — 👍 21 🔁 5 💬 0 📌 0
07.01.2026 11:13 — 👍 30 🔁 6 💬 0 📌 0
I probably shouldn't have used this photo with the obvious hand trouble, but I thought it was a good picture!
01.01.2026 13:46 — 👍 1 🔁 0 💬 0 📌 04. Ambiguity (i.e. - the positioning of hands - facing the camera, angles of fingers - causes symmetry problems).
5. Generative models don’t understand anatomy. Even the most up-to-date rendering engines cause trouble.
1. Hands are anatomically insane.
2. Training data is messy and inconsistent.
3. Hands lose the attention battle (i.e. - the renderer give more weight to shadows, faces, lighting, etc, than hands).
I agree - it seems to love putting an extra finger, or twisty fingers, or whatever! I searched around about this - after all, this has been a complaint since AI began.
Hands are a perfect storm of complexity, ambiguity, and data scarcity. Even in 2025, top-tier models still glitch on them.
01.01.2026 13:38 — 👍 61 🔁 9 💬 0 📌 0
25.12.2025 19:11 — 👍 29 🔁 4 💬 0 📌 0
