First camera-control LoRA for Qwen-Image-2.1 reshoots any photo from 72 addressable angles
_akhaliq · x · 2026-10-07
akhaliq released the first camera-control LoRA for Qwen-Image-2.1, letting you feed a photo and regenerate the same subject from a controlled azimuth/elevation.
- Trigger <mva>, works in ComfyUI / diffusers at strength 0.8–1.0
- Pose dictionary: 12 azimuths (30° steps) × 4 elevations plus close-up variants — 72 addressable framings
- Trained on 5,028 Dome-Objaverse pairs; a v2 (13k pairs, rank 64, bf16 base) is training to fix identity fidelity on characters and out-of-distribution photos, expected within a day
- Apache-2.0, v1 recommended checkpoint unchanged
Related event: Camera-Control LoRA for Qwen-Image-2.1 Reshoots Any Photo from 72 Angles(2 posts)→
More from Multimodal
- VEDA Sparse Attention cuts MiniMax H3 video gen time in half in ComfyUI with no visible quality loss — robomar_ai_art · 2026-10-07
- a16z consumer AI overview turned into a 3-minute video with Manus — parker_lyman · 2026-10-07
- Blogger says Opus-generated video matches museum-grade immersive work that once cost $70K teams — oran_ge · 2026-10-07
- Super robot anime made with Kling 4 Flash looks straight out of a studio — Forsaken_Stuff_Ai · 2026-10-07
- Dev builds ComfyUI node pack for projection mapping, 3D region prompts and motion design — Puzzled_Parking2556 · 2026-10-07
- Full Nano Banana 2.1 Prompt for Cinematic Posters That Keep Your Face and Copy the Layout — CodeByPoonam · 2026-10-07