dotsimulate
diff --git a/‎.gitignore‎
Lines changed: 9 additions & 0 deletions b/‎.gitignore‎
Lines changed: 9 additions & 0 deletions
diff --git a/‎configs/_preprocessor_reference.yaml.example‎
Lines changed: 131 additions & 0 deletions b/‎configs/_preprocessor_reference.yaml.example‎
Lines changed: 131 additions & 0 deletions
diff --git a/‎configs/prompt_blending_demo.yaml.example‎
Lines changed: 54 additions & 0 deletions b/‎configs/prompt_blending_demo.yaml.example‎
Lines changed: 54 additions & 0 deletions
diff --git a/‎configs/sd15_canny_depth.yaml.example‎
Lines changed: 45 additions & 0 deletions b/‎configs/sd15_canny_depth.yaml.example‎
Lines changed: 45 additions & 0 deletions
diff --git a/‎configs/sd15_depth_trt_example.yaml.example‎
Lines changed: 42 additions & 0 deletions b/‎configs/sd15_depth_trt_example.yaml.example‎
Lines changed: 42 additions & 0 deletions
diff --git a/‎configs/sd15_tile.yaml.example‎
Lines changed: 28 additions & 0 deletions b/‎configs/sd15_tile.yaml.example‎
Lines changed: 28 additions & 0 deletions
diff --git a/‎configs/sdturbo_color.yaml.example‎
Lines changed: 33 additions & 0 deletions b/‎configs/sdturbo_color.yaml.example‎
Lines changed: 33 additions & 0 deletions
@@ -1,4 +1,5 @@
 # https://github.com/github/gitignore/blob/main/Python.gitignore
+notes/
 
 .vscode/
 engines/
@@ -206,3 +207,11 @@ __pycache__/
 
 models/RealESR*
 *.safetensors
+*.onnx
+
+# Ignore user config files
+configs/*.yaml
+configs/*.yml
+
+images/outputs/*
+images/inputs/*
@@ -0,0 +1,131 @@
+# StreamDiffusion Preprocessor Reference
+# Copy and modify the sections you need for your configuration
+
+# =============================================================================
+# PREPROCESSOR INTERFACES  
+# =============================================================================
+
+# Basic Preprocessors  
+# ---------------------------------------------------
+
+# Canny Edge Detection
+preprocessor: "canny"
+preprocessor_params:
+  low_threshold: 100    # int, 0-255, lower values detect more edges
+  high_threshold: 200   # int, 0-255, should be 2-3x higher than low_threshold
+
+# Passthrough (no processing - for tile/reference ControlNets)
+preprocessor: "passthrough"
+preprocessor_params:
+  image_resolution: 512  # int, 256-1024, output resolution
+
+# CPU Depth Estimation (requires transformers)
+preprocessor: "depth"
+preprocessor_params:
+  model_name: "Intel/dpt-large"    # str, HuggingFace model name
+  detect_resolution: 512           # int, resolution for detection
+  image_resolution: 512            # int, final output resolution
+
+# CPU Pose Detection (requires controlnet_aux or fallback)
+preprocessor: "openpose"
+preprocessor_params:
+  detect_resolution: 512     # int, resolution for detection  
+  image_resolution: 512      # int, final output resolution
+  include_hands: false       # bool, include hand keypoints
+  include_face: false        # bool, include face keypoints
+
+# Line Art Detection (requires controlnet_aux)
+preprocessor: "lineart"
+preprocessor_params:
+  detect_resolution: 512     # int, resolution for detection
+  image_resolution: 512      # int, final output resolution
+  coarse: true              # bool, use coarse line art detection
+  anime_style: false        # bool, use anime-style line art
+
+# Standard Line Art (traditional computer vision)
+preprocessor: "standard_lineart"
+preprocessor_params:
+  detect_resolution: 512        # int, resolution for detection
+  image_resolution: 512         # int, final output resolution  
+  gaussian_sigma: 6.0          # float, Gaussian blur sigma
+  intensity_threshold: 8       # int, intensity calculation threshold
+
+# Soft Edge Detection (real-time optimized)
+preprocessor: "soft_edge"
+preprocessor_params:
+  image_resolution: 512         # int, 256-1024, final output resolution (512 recommended for optimal performance)
+  strength: 1.0                # float, 0.1-2.0, edge detection strength
+  soft_threshold: 0.5          # float, 0.1-0.9, controls edge softness
+  multi_scale: true            # bool, use multi-scale edge detection for better quality
+  gaussian_sigma: 1.0          # float, 0.5-3.0, pre-smoothing sigma
+
+# Advanced Preprocessors (Require additional setup)
+# --------------------------------------------------
+
+# TensorRT Depth Estimation (requires TensorRT engine)
+preprocessor: "depth_tensorrt"
+preprocessor_params:
+  engine_path: "C:\\path\\to\\depth_anything_v2_vits-fp16.engine"  # str, REQUIRED: absolute path to .engine file
+  detect_resolution: 518     # int, 256-1024, must match engine input size
+  image_resolution: 512      # int, 256-1024, final output resolution
+
+# TensorRT Pose Detection (requires TensorRT engine)
+preprocessor: "pose_tensorrt"
+preprocessor_params:
+  engine_path: "C:\\path\\to\\yolo_nas_pose_l_0.8-fp16.engine"  # str, REQUIRED: absolute path to .engine file
+  detect_resolution: 640     # int, 320-1024, must match engine input size  
+  image_resolution: 512      # int, 256-1024, final output resolution
+
+# MediaPipe Pose Detection (requires mediapipe)
+preprocessor: "mediapipe_pose"
+preprocessor_params:
+  detect_resolution: 512              # int, resolution for detection
+  image_resolution: 512               # int, final output resolution
+  min_detection_confidence: 0.5       # float, 0.0-1.0, minimum detection confidence
+  min_tracking_confidence: 0.5        # float, 0.0-1.0, minimum tracking confidence
+  model_complexity: 1                 # int, 0-2, MediaPipe model complexity
+  static_image_mode: true             # bool, treat each image independently
+  draw_hands: true                    # bool, draw hand poses
+  draw_face: false                    # bool, draw face landmarks
+  line_thickness: 2                   # int, thickness of skeleton lines
+  circle_radius: 4                    # int, radius of joint circles
+  confidence_threshold: 0.3           # float, 0.0-1.0, minimum confidence for rendering
+  enable_smoothing: true              # bool, enable temporal smoothing
+  smoothing_factor: 0.7               # float, 0.0-1.0, smoothing strength
+
+# =============================================================================
+# AVAILABLE CONTROLNET MODELS (include but are not limited to)
+# =============================================================================
+
+# Stable Diffusion 1.5 Models
+sd15_models_include_but_not_limited_to:
+  canny: "lllyasviel/control_v11p_sd15_canny"
+  depth: "lllyasviel/control_v11f1p_sd15_depth"
+  openpose: "lllyasviel/control_v11p_sd15_openpose"
+  lineart: "lllyasviel/control_v11p_sd15_lineart"
+  tile: "lllyasviel/control_v11f1e_sd15_tile"
+  scribble: "lllyasviel/control_v11p_sd15_scribble"
+  seg: "lllyasviel/control_v11p_sd15_seg"
+  normal: "lllyasviel/control_v11p_sd15_normalbae"
+  mlsd: "lllyasviel/control_v11p_sd15_mlsd"
+  hed: "lllyasviel/control_v11p_sd15_hed"  # Compatible with soft_edge preprocessor
+
+# Stable Diffusion 2.1 Models  
+sd21_models_include_but_not_limited_to:
+  canny: "thibaud/controlnet-sd21-canny-diffusers"
+  depth: "thibaud/controlnet-sd21-depth-diffusers"
+  openpose: "thibaud/controlnet-sd21-openpose-diffusers"
+  lineart: "thibaud/controlnet-sd21-lineart-diffusers"
+  scribble: "thibaud/controlnet-sd21-scribble-diffusers"
+  hed: "thibaud/controlnet-sd21-hed-diffusers"
+  zoedepth: "thibaud/controlnet-sd21-zoedepth-diffusers"
+
+# SDXL Models
+sdxl_models_include_but_not_limited_to:
+  canny: "diffusers/controlnet-canny-sdxl-1.0"
+  depth: "diffusers/controlnet-depth-sdxl-1.0"
+  openpose: "thibaud/controlnet-openpose-sdxl-1.0"
+  # Note: soft_edge preprocessor compatible with HED-based ControlNet models
+
+# Note: Many more ControlNet models exist on HuggingFace Hub.
+# Search for "controlnet" to find additional models for your specific needs. 
@@ -0,0 +1,54 @@
+# StreamDiffusion Configuration for Prompt & Seed Blending Demo
+# Simple img2img setup without ControlNets for demonstrating prompt and seed blending
+
+model_id: "KBlueLeaf/kohaku-v2.1"
+t_index_list: [16, 32]
+width: 512
+height: 512
+device: "cuda"
+dtype: "float16"
+mode: img2img
+
+# Generation parameters (base - will be overridden by blending)
+prompt: "a waifu girl cute"
+negative_prompt: "blurry, low quality, ugly"
+guidance_scale: 1.2
+num_inference_steps: 50
+
+# StreamDiffusion parameters
+use_denoising_batch: true
+delta: 0.7
+frame_buffer_size: 1
+
+# Pipeline configuration
+pipeline_type: "sd1.5"
+use_lcm_lora: true
+use_tiny_vae: true
+acceleration: "xformers"  # Use xformers instead of tensorrt for easier setup
+cfg_type: "self"
+seed: 42
+
+# Warmup iterations for performance
+warmup: 5
+
+# Prompt blending configuration
+# This will override the single 'prompt' above
+# prompt_blending:
+#   prompt_list:
+#     - ["a waifu girl cute", 1.0]
+#     - ["a demon from hell", 0.0]
+#   interpolation_method: "slerp"  # or "linear"
+#   enable_caching: true
+
+# Seed blending configuration
+# This enables blending between different noise patterns
+# for added visual variety alongside prompt blending
+seed_blending:
+  seed_list:
+    - [42, 1.0]      # Stable, controlled generation
+    - [999, 0.0]     # More chaotic, varied generation
+  interpolation_method: "linear"  # or "slerp"
+  enable_caching: true
+
+# No ControlNets for this demo
+# controlnets: [] 
@@ -0,0 +1,45 @@
+model_id: "stabilityai/sd-turbo"
+t_index_list: [0,16]
+width: 512
+height: 512
+device: "cuda"
+dtype: "float16"
+
+# Generation parameters
+prompt: "an anime render of a girl with purple hair, masterpiece"
+negative_prompt: "blurry, low quality, flat, 2d"
+guidance_scale: 1.1
+num_inference_steps: 50
+
+# Temporal consistency parameters
+frame_buffer_size: 1
+delta: 0.7
+
+# Advanced parameters
+use_lcm_lora: false
+use_tiny_vae: true
+acceleration: "tensorrt"
+cfg_type: "self"
+seed: 789
+
+# ControlNet configuration with TensorRT Depth Anything
+controlnets:
+  - model_id: "thibaud/controlnet-sd21-depth-diffusers"
+    conditioning_scale: 0.5
+    preprocessor: "depth_tensorrt"
+    preprocessor_params:
+      engine_path: "C:\\_dev\\comfy\\ComfyUI\\models\\tensorrt\\depth-anything\\depth_anything_vits14-fp16.engine"
+      detect_resolution: 518
+      image_resolution: 512
+    enabled: true 
+
+  - model_id: "thibaud/controlnet-sd21-canny-diffusers"
+    conditioning_scale: 0.5
+    preprocessor: "canny"
+    preprocessor_params:
+      low_threshold: 50
+      high_threshold: 100
+    control_image_path: null
+    enabled: true
+    control_guidance_start: 0.0
+    control_guidance_end: 1.0 
@@ -0,0 +1,42 @@
+model_id: "KBlueLeaf/kohaku-v2.1"
+t_index_list: [16, 32]
+width: 512
+height: 512
+device: "cuda"
+dtype: "float16"
+
+# Generation parameters
+prompt: "an anime render of a girl with purple hair, masterpiece"
+negative_prompt: "blurry, low quality, flat, 2d"
+guidance_scale: 1.1
+num_inference_steps: 50
+
+use_denoising_batch: true
+delta: 0.7
+frame_buffer_size: 1
+
+# Advanced parameters
+use_lcm_lora: true
+use_tiny_vae: true
+acceleration: "xformers"
+cfg_type: "self"
+seed: 789
+
+# ControlNet configuration with TensorRT Depth Anything
+controlnets:
+  - model_id: "lllyasviel/control_v11f1p_sd15_depth"
+    conditioning_scale: 0.28
+    preprocessor: "depth_tensorrt"
+    preprocessor_params:
+      engine_path: "C:\\_dev\\comfy\\ComfyUI\\models\\tensorrt\\depth-anything\\v2_depth_anything_v2_vits-fp16.engine"
+      detect_resolution: 518
+      image_resolution: 512
+    enabled: true 
+
+  - model_id: "lllyasviel/control_v11p_sd15_canny"
+    conditioning_scale: 0.29
+    preprocessor: "canny"
+    preprocessor_params:
+      low_threshold: 100
+      high_threshold: 200
+    enabled: true 
@@ -0,0 +1,28 @@
+model_id: "KBlueLeaf/kohaku-v2.1"
+t_index_list: [0,16]
+width: 512
+height: 512
+device: "cuda"
+dtype: "float16"
+
+# Generation parameters
+prompt: "masterpiece, high quality, detailed"
+negative_prompt: "blurry, low quality, distorted"
+guidance_scale: 1.1
+num_inference_steps: 50
+
+# Advanced parameters
+use_lcm_lora: true
+use_tiny_vae: true
+acceleration: "xformers"
+cfg_type: "self"
+seed: 456
+
+# ControlNet configuration with Tile ControlNet
+controlnets:
+  - model_id: "lllyasviel/control_v11f1e_sd15_tile"
+    conditioning_scale: 0.2
+    preprocessor: "passthrough"
+    preprocessor_params:
+      image_resolution: 512
+    enabled: true 
@@ -0,0 +1,33 @@
+model_id: "stabilityai/sd-turbo"
+t_index_list: [0,16]
+width: 512
+height: 512
+device: "cuda"
+dtype: "float16"
+
+# Generation parameters
+prompt: "an anime render of a girl with purple hair, masterpiece"
+negative_prompt: "blurry, low quality, flat, 2d"
+guidance_scale: 1.1
+num_inference_steps: 50
+
+# Temporal consistency parameters
+frame_buffer_size: 1
+use_denoising_batch: true
+delta: 0.7
+
+# Advanced parameters
+use_lcm_lora: true
+use_tiny_vae: true
+acceleration: "tensorrt"
+cfg_type: "self"
+seed: 789
+
+# ControlNet configuration with TensorRT Depth Anything
+controlnets:
+  - model_id: "thibaud/controlnet-sd21-color-diffusers"
+    conditioning_scale: 0.2
+    preprocessor: "passthrough"
+    preprocessor_params:
+      image_resolution: 512
+    enabled: true