blender-cameras
Set up Blender cameras with cinematic intent — focal length, depth of field (f-stop / focus object), composition (rule of thirds, leading lines), animated cameras (orbit, dolly, push-in), tracking constraints. Use whenever the user asks to "set up the camera", "frame the shot", "make it look cinematic / hero / portrait / wide-angle / telephoto", "add depth of field", "orbit the camera", or any composition/framing request. Make sure to use this skill even if the user does not say "camera" — also covers "hero shot", "close-up", "from above", "shallow focus", "85mm portrait look".
CheshireJCat · 已有静态检查记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add CheshireJCat/blender --skill blender-cameras技能说明
name: blender-cameras description: Set up Blender cameras with cinematic intent — focal length, depth of field (f-stop / focus object), composition (rule of thirds, leading lines), animated cameras (orbit, dolly, push-in), tracking constraints. Use whenever the user asks to "set up the camera", "frame the shot", "make it look cinematic / hero / portrait / wide-angle / telephoto", "add depth of field", "orbit the camera", or any composition/framing request. Make sure to use this skill even if the user does not say "camera" — also covers "hero shot", "close-up", "from above", "shallow focus", "85mm portrait look". when_to_use: Any camera placement, framing, focal length, DoF, or animated camera setup in Blender. allowed-tools: Read Bash blender_python blender_scene_info blender_scene_info
Blender Cameras
Set up cameras with the same decisions a real cinematographer makes: focal length for feel, f-stop for focus, composition for storytelling.
Focal length cheat sheet
| Length | Feel | Use |
|---|---|---|
| 14–24mm | Very wide, distorted | Architecture, claustrophobic interiors, exaggerated perspective |
| 28–35mm | Wide, "documentary" | Establishing shots, environments |
| 50mm | Neutral (≈ human eye) | Default storytelling |
| 85mm | Short telephoto | Portraits, character close-ups (flattering) |
| 100–135mm | Telephoto | Hero product shots, isolated subjects |
| 200mm+ | Long tele | Wildlife, surveillance look, heavy compression |
Quick rule: 85mm for intimacy, 24mm for spectacle, 50mm for neutral.
Aperture / f-stop
| f-stop | DoF | Use |
|---|---|---|
| f/1.2–2.0 | Razor thin | Hero portraits, dreamy |
| f/2.8 | Shallow | Standard portrait |
| f/4 | Moderate | Two subjects in frame |
| f/5.6–8 | Medium-deep | Group portraits, environments |
| f/11+ | Very deep | Landscape, "everything sharp" |
Recipes
Recipe 0 — Bbox-aware hero camera (preferred for orchestrator chains)
Use this when you have a specific subject. Computes the subject's bounding box, places camera at a distance that fits the subject in ~80% of the frame vertically, and aims via Track-To.
import bpy, math
from mathutils import Vector
# Choose subject — all meshes named GEO-* by default, or pass a specific list
subject_meshes = [o for o in bpy.data.objects if o.type == 'MESH' and o.name.startswith('GEO-')]
if not subject_meshes:
raise RuntimeError("No subject meshes found (looking for GEO- prefix)")
# World-space bbox
deps = bpy.context.evaluated_depsgraph_get()
all_verts = []
for o in subject_meshes:
eo = o.evaluated_get(deps); em = eo.to_mesh()
for v in em.vertices:
all_verts.append(o.matrix_world @ v.co)
eo.to_mesh_clear()
xs = [v.x for v in all_verts]; ys = [v.y for v in all_verts]; zs = [v.z for v in all_verts]
center = Vector(((min(xs)+max(xs))/2, (min(ys)+max(ys))/2, (min(zs)+max(zs))/2))
height = max(zs) - min(zs)
width = max(xs) - min(xs)
biggest = max(height, width)
# Frame fit: at distance D, vertical frame = D × (sensor_h / focal). Solve for D.
focal_mm = 60 # 60mm gives a flattering not-too-wide hero shot
sensor_h_mm = 24 # full-frame
frame_per_meter = sensor_h_mm / focal_mm # 0.4 m vertical frame per metre of distance
target_fill = 0.80
camera_distance = biggest / (frame_per_meter * target_fill)
# Camera positioned in front (negative Y) with slight X offset for a 3/4 angle
cam_pos = Vector((center.x + camera_distance * 0.3, center.y - camera_distance, center.z))
# Empty for tracking
empty_name = 'Empty-camera_target'
empty = bpy.data.objects.get(empty_name) or bpy.data.objects.new(empty_name, None)
if empty.name not in [o.name for o in bpy.context.collection.objects]:
bpy.context.collection.objects.link(empty)
empty.location = center
# Camera
cam_data = bpy.data.cameras.new('CAM-hero')
cam_data.lens = focal_mm
cam_data.dof.use_dof = True
cam_data.dof.aperture_fstop = 4.0
cam_data.dof.focus_object = subject_meshes[0] # focus on first/main subject
cam = bpy.data.objects.new('CAM-hero', cam_data)
bpy.context.collection.objects.link(cam)
cam.location = cam_pos
track = cam.constraints.new('TRACK_TO')
track.target = empty
track.track_axis = 'TRACK_NEGATIVE_Z'
track.up_axis = 'UP_Y'
bpy.context.scene.camera = cam
print(f"camera:bbox_aware center={tuple(round(v,2) for v in center)} dist={camera_distance:.2f}m focal={focal_mm}mm")
For elongated vertical subjects (sword, flag, candle): biggest dimension is height; the framing math fits height to 80% of vertical frame, which is what you want.
For wide horizontal subjects (car, table): biggest is width; it fits width to 80% of vertical frame too which over-zooms — for those, swap to frame_per_meter_h = (sensor_h_mm * aspect_ratio) / focal_mm or adjust target_fill down.
Recipe 1 — Hero portrait camera (85mm + shallow DoF)
import bpy, math
subject = bpy.data.objects.get('GEO-subject') # change to your subject
# Camera
cam_data = bpy.data.cameras.new('CAM-hero')
cam = bpy.data.objects.new('CAM-hero', cam_data)
bpy.context.collection.objects.link(cam)
bpy.context.scene.camera = cam
cam.location = (3, -4, 1.6)
cam_data.lens = 85
cam_data.sensor_width = 36
# Depth of field
cam_data.dof.use_dof = True
cam_data.dof.aperture_fstop = 2.8
if subject:
cam_data.dof.focus_object = subject
# Track-to constraint (auto-aim at subject)
if subject:
track = cam.constraints.new('TRACK_TO')
track.target = subject
track.track_axis = 'TRACK_NEGATIVE_Z'
track.up_axis = 'UP_Y'
print('camera:CAM-hero set')
Recipe 2 — Wide environmental establishing shot (24mm)
import bpy, math
cam_data = bpy.data.cameras.new('CAM-establish')
cam = bpy.data.objects.new('CAM-establish', cam_data)
bpy.context.collection.objects.link(cam)
bpy.context.scene.camera = cam
cam.location = (8, -10, 2.5)
cam.rotation_euler = (math.radians(80), 0, math.radians(35))
cam_data.lens = 24
cam_data.sensor_width = 36
cam_data.dof.use_do
阅读完整来源 ↗