计划执行与检查节点
executing-plans
按既定实施计划完成工作,保持过程检查和反馈。
blender-render-automation【效率与协作】Automate Blender rendering from the command line. Use when the user wants to set up renders, batch render scenes, configure Cycles or EEVEE, set up cameras and lights, render animations, create materials and shaders, or build a render pipeline with Blender Python scripting.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Automate Blender's rendering pipeline from the terminal. Configure render engines (Cycles/EEVEE), set up cameras and lighting, create materials, and batch render scenes or animations — all headlessly via Python scripts.
import bpy
scene = bpy.context.scene
# Cycles (ray-traced, production quality)
scene.render.engine = 'CYCLES'
cycles = scene.cycles
cycles.samples = 256
cycles.use_denoising = True
cycles.denoiser = 'OPENIMAGEDENOISE'
cycles.device = 'GPU'
prefs = bpy.context.preferences.addons['cycles'].preferences
prefs.compute_device_type = 'OPTIX' # or 'CUDA', 'HIP', 'ONEAPI', 'METAL'
prefs.get_devices()
for device in prefs.devices:
device.use = device.type != 'CPU' # GPU only; leave CPU on to add it as a helper
# EEVEE (fast, real-time). The engine id changed: 'BLENDER_EEVEE_NEXT' in
# Blender 4.2-4.5, 'BLENDER_EEVEE' again from 5.0. Reading the accepted
# values avoids hardcoding either.
ids = [e.identifier for e in scene.render.bl_rna.properties['engine'].enum_items]
scene.render.engine = 'BLENDER_EEVEE' if 'BLENDER_EEVEE' in ids else 'BLENDER_EEVEE_NEXT'
scene.eevee.taa_render_samples = 64
blender -E help lists the engines your build accepts. EEVEE in background mode
still needs a working OpenGL context; on a server without a display use Cycles,
or run Blender under a virtual display.
render = bpy.context.scene.render
render.resolution_x = 1920
render.resolution_y = 1080
render.resolution_percentage = 100
render.image_settings.file_format = 'PNG' # PNG, JPEG, OPEN_EXR, TIFF
render.image_settings.color_mode = 'RGBA'
render.film_transparent = True # transparent background
Blender 5.0 added image_settings.media_type ('IMAGE', 'MULTI_LAYER_IMAGE',
'VIDEO'). Set it before file_format, otherwise choosing 'FFMPEG' fails
(see the video example below). The attribute does not exist in 4.x.
import math
from mathutils import Vector
# Camera
bpy.ops.object.camera_add(location=(7, -6, 5))
camera = bpy.context.active_object
target = Vector((0, 0, 1))
direction = target - camera.location
camera.rotation_euler = direction.to_track_quat('-Z', 'Y').to_euler()
cam_data = camera.data
cam_data.lens = 50
cam_data.dof.use_dof = True
cam_data.dof.focus_distance = 5
cam_data.dof.aperture_fstop = 2.8
bpy.context.scene.camera = camera
# Track-to constraint (auto-aim)
track = camera.constraints.new(type='TRACK_TO')
track.target = bpy.data.objects["MySubject"]
# Area light
bpy.ops.object.light_add(type='AREA', location=(0, -4, 3))
area = bpy.context.active_object
area.data.energy = 500
area.data.size = 2
# HDRI environment lighting
world = bpy.context.scene.world or bpy.data.worlds.new("World")
bpy.context.scene.world = world
world.use_nodes = True # deprecated no-op in 5.0 (nodes are always on), required in 4.x
nodes = world.node_tree.nodes
links = world.node_tree.links
nodes.clear()
bg = nodes.new('ShaderNodeBackground')
env = nodes.new('ShaderNodeTexEnvironment')
output = nodes.new('ShaderNodeOutputWorld')
env.image = bpy.data.images.load("/home/render/hdri/studio_small_09_2k.hdr")
links.new(env.outputs['Color'], bg.inputs['Color'])
links.new(bg.outputs['Background'], output.inputs['Surface'])
def create_pbr_material(name, color, metallic=0.0, roughness=0.5):
mat = bpy.data.materials.new(name)
mat.use_nodes = True # deprecated no-op in 5.0, required in 4.x
# look the node up by type: names are localized on non-English UIs
bsdf = next(n for n in mat.node_tree.nodes if n.type == 'BSDF_PRINCIPLED')
bsdf.inputs['Base Color'].default_value = (*color, 1)
bsdf.inputs['Metallic'].default_value = metallic
bsdf.inputs['Roughness'].default_value = roughness
return mat
def create_glass_material(name, color=(1, 1, 1), ior=1.45):
mat = bpy.data.materials.new(name)
mat.use_nodes = True
bsdf = next(n for n in mat.node_tree.nodes if n.type == 'BSDF_PRINCIPLED')
bsdf.inputs['Base Color'].default_value = (*color, 1)
bsdf.inputs['Transmission Weight'].default_value = 1.0
bsdf.inputs['Roughness'].default_value = 0.0
bsdf.inputs['IOR'].default_value = ior
return mat
obj = bpy.data.objects["MyCube"]
obj.data.materials.append(create_pbr_material("BlueMetal", (0.1, 0.3, 0.8), metallic=1.0, roughness=0.2))
# Single frame
scene.render.filepath = "//renders/hero.png" # // = folder of the .blend
bpy.ops.render.render(write_still=True)
# Animation as image sequence
scene.frame_start = 1
scene.frame_end = 250
scene.render.fps = 24
scene.render.filepath = "//renders/anim/frame_" # frames get a number suffix: frame_0001.png
bpy.ops.render.render(animation=True)
# Animation as video
scene.render.filepath = "//renders/animation.mp4"
if hasattr(scene.render.image_settings, 'media_type'): # Blender 5.0+
scene.render.image_settings.media_type = 'VIDEO'
scene.render.image_settings.file_format = 'FFMPEG'
scene.render.ffmpeg.format = 'MPEG4'
scene.render.ffmpeg.codec = 'H264'
bpy.ops.render.render(animation=True)
CLI (no script needed). Arguments run in the order given, so put the frame/animation flag last:
blender -b scene.blend -E CYCLES -o //renders/frame_#### -F PNG -f 1
blender -b scene.blend -s 1 -e 100 -o //renders/shot_#### -a
blender -b scene.blend --python setup_render.py -f 1 -- --cycles-device OPTIX
# characters set the frame-number padding; -F overrides the saved format.
Arguments after a lone -- are ignored by Blender, so your own script can read
them from sys.argv and Cycles reads --cycles-device.
import os
output_dir = bpy.path.abspath("//renders/cameras")
os.makedirs(output_dir, exist_ok=True)
scene = bpy.context.scene
for cam in [obj for obj in bpy.data.objects if obj.type == 'CAMERA']:
scene.camera = cam
scene.render.filepath = os.path.join(output_dir, f"{cam.name}.png")
bpy.ops.render.render(write_still=True)
User request: "Set up a clean studio render for a 3D product"
Output: Script that clears the scene, imports the product OBJ, creates a white backdrop with PBR material, sets up three-point lighting (key area light, fill, rim), adds a camera with Track-To constraint aimed at the product, configures Cycles at 128 samples with denoising, transparent background (RGBA), and renders at 2000x2000.
Run it with blender -b --factory-startup --python product_shot.py --python-exit-code 1; the result is renders/product_2000.png with a transparent background.
User request: "Render a 360-degree turntable of my model — 36 frames"
Output: Script that creates a camera, loops 36 steps around the model at equal angular intervals using cos/sin, renders each frame with Cycles + denoising to a numbered PNG sequence, then provides the ffmpeg command to assemble into an MP4.
blender -b turntable.blend --python turntable.py
ffmpeg -framerate 12 -i renders/turntable/frame_%04d.png -c:v libx264 -pix_fmt yuv420p turntable.mp4
The result is 36 PNGs (frame_0001.png to frame_0036.png) and a 3-second MP4.
--background (-b) for headless rendering. Add --python-exit-code 1 so a script error fails your CI job instead of exiting 0.--disable-autoexec.--factory-startup to keep user preferences from changing results.prefs.get_devices() after setting compute_device_type.film_transparent = True and RGBA for renders needing transparent backgrounds.Transmission before Blender 4.0).Z is now Depth, DiffCol is Diffuse Color) and moved the compositor to scene.compositing_node_group; scripts that touch either need a version check.