메뉴
Headless Blender 5 rendering from Docker with Cycles CPU, material creation patterns, and S3 delivery. Use when rendering 3D scenes via Blender in container.
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/arm2arm/AstroAgentAssistant --skill blender-rendering명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
SOC 직업 분류 기준
SKILL.md 표시 중
| name | blender-rendering |
| description | Headless Blender 5 rendering from Docker with Cycles CPU, material creation patterns, and S3 delivery. Use when rendering 3D scenes via Blender in container. |
| tags | ["blender","rendering","docker","cycles","python"] |
blender-mcp-headless Docker imageDocker image: blender-mcp-headless:latest (built on linuxserver/blender arm64)
Blender version: 5.0.1
Render device: CPU only (GPU/GPU-EGL fails in headless mode — produces all-black frames)
docker run --rm \
-v /path/to/script.py:/scripts/script.py:ro \
-v /path/to/output:/output:rw \
blender-mcp-headless:latest \
blender -b --python /scripts/script.py
-b = background mode (no GUI)/output/ for files to persist/usr/bin/blender (via python3.3 symlink)bpy.ops.wm.read_factory_setting → bpy.ops.wm.read_factory_settings (plural "settings")ShaderNodeOutputMaterial.outputs["BSDF"] → outputs remain "BSDF" for Principled/GlossyShaderNodeBackground.outputs["BSDF"] → outputs renamed to "Background"Vector.to_track_quaternion() → still works but prefer constraints.new(type='TRACK_TO')Light.data.rotation_euler → use Object.rotation_euler instead (on the light object, not data)Material.use_nodes → deprecated, will be removed in 6.0 (still works, just warns)Principled BSDF → "Specular" and "Sheen" inputs may not exist — always check with:
input_names = [inp.name for inp in principled.inputs]
if "Specular" in input_names:
principled.inputs["Specular"].default_value = value
'BLENDER_EEVEE_NEXT' doesn't exist — use 'BLENDER_EEVEE' or 'CYCLES'scene.cycles.device = 'GPU' in headless Docker produces all-black renders (EGL context issues). Always use 'CPU'.BLENDER_EEVEE requires a display/Wayland socket — may fail in some headless configs. Cycles CPU is most reliable.def make_mirror_mat():
mat = bpy.data.materials.new("Mirror")
mat.use_nodes = True
nt = mat.node_tree
nt.nodes.clear()
out = nt.nodes.new("ShaderNodeOutputMaterial")
glossy = nt.nodes.new("ShaderNodeBsdfGlossy")
glossy.inputs["Color"].default_value = (1.0, 1.0, 1.0, 1.0)
glossy.inputs["Roughness"].default_value = 0.005 # not 0.0 — avoids delta issues
nt.links.new(glossy.outputs["BSDF"], out.inputs["Surface"])
return mat
def safe_principled(name, base_color, roughness=0.5, metallic=0.0, specular=0.5):
mat = bpy.data.materials.new(name)
mat.use_nodes = True
nt = mat.node_tree
nodes = nt.nodes
links = nt.links
nodes.clear()
out = nodes.new("ShaderNodeOutputMaterial")
out.location = (300, 0)
principled = nodes.new("ShaderNodeBsdfPrincipled")
principled.inputs["Base Color"].default_value = base_color
principled.inputs["Roughness"].default_value = roughness
principled.inputs["Metallic"].default_value = metallic
input_names = [inp.name for inp in principled.inputs]
if "Specular" in input_names:
principled.inputs["Specular"].default_value = specular
links.new(principled.outputs["BSDF"], out.inputs["Surface"])
return mat
def setup_world(color=(0.15, 0.15, 0.18, 1.0), strength=1.0):
scene.world = bpy.data.worlds.new("World")
scene.world.use_nodes = True
nodes = scene.world.node_tree.nodes
links = scene.world.node_tree.links
nodes.clear()
w_out = nodes.new("ShaderNodeOutputWorld")
w_bg = nodes.new("ShaderNodeBackground")
w_bg.inputs[0].default_value = color
w_bg.inputs[1].default_value = strength
links.new(w_bg.outputs["Background"], w_out.inputs["Surface"])
bpy.ops.object.camera_add()
camera = bpy.context.object
camera.location = Vector((x, y, z)) # top-down 45° angle
target = bpy.data.objects.new("CamTarget", None)
bpy.context.collection.objects.link(target)
target.location = Vector((look_x, look_y, look_z))
constraint = camera.constraints.new(type='TRACK_TO')
constraint.target = target
constraint.track_axis = 'TRACK_NEGATIVE_Z'
constraint.up_axis = 'UP_Y'
scene.camera = camera
# Key light
bpy.ops.object_light_add(type='SUN')
key = bpy.context.object
key.data.energy = 10.0
key.rotation_euler = (math.radians(35), 0, math.radians(45))
# Fill light
bpy.ops.object_light_add(type='SUN')
fill = bpy.context.object
fill.data.energy = 4.0
fill.rotation_euler = (math.radians(55), 0, math.radians(-135))
# Rim light
bpy.ops.object_light_add(type='SUN')
rim = bpy.context.object
rim.data.energy = 6.0
rim.rotation_euler = (math.radians(20), 0, math.radians(200))
Blender's primitive_plane_add() creates a plane in the XY-plane (normal = +Z). For a vertical mirror facing +Y:
bpy.ops.mesh.primitive_plane_add()
mirror = bpy.context.object
mirror.location = (0, -3.0, 0.5) # mirror position
mirror.rotation_euler = (math.radians(90), 0, 0) # rotate to YZ plane, facing +Y
mirror.scale = (width * 0.48, 1.0, height * 0.48) # scale X and Z (not Y)
glossy_bounces >= 8 in Cycles settingsscene.render.engine = 'CYCLES'
scene.cycles.device = 'CPU' # Always CPU in headless Docker
scene.cycles.samples = 256 # or 512/1024 for final
scene.cycles.use_denoising = True
scene.cycles.max_bounces = 12
scene.cycles.glossy_bounces = 8 # Critical for mirror reflections
scene.cycles.diffuse_bounces = 8
scene.cycles.transmission_bounces = 8
scene.render.resolution_x = 1920
scene.render.resolution_y = 1080
scene.render.image_settings.file_format = 'PNG'
scene.render.image_settings.color_mode = 'RGBA'
# After render, upload to S3 and get Telegram URL
python3 ~/.hermes/scripts/s3_media_upload.py /output/render.png
/home/hermes/projects/manim-tutorials/scripts//home/hermes/projects/manim-tutorials/output/blender-mcp-headless:latest