Merge pull request #118 from benrugg/pullrequests/AyaanShah2204/main

Pullrequests/ayaan shah2204/main

__init__.py

@@ -2,7 +2,7 @@
     "name": "AI Render - Stable Diffusion in Blender",
     "description": "Create amazing images using Stable Diffusion AI",
     "author": "Ben Rugg",
-    "version": (0, 8, 0),
+    "version": (0, 9, 0),
     "blender": (3, 0, 0),
     "location": "Render Properties > AI Render",
     "warning": "",

operators.py

@@ -540,6 +540,191 @@ def sd_upscale(scene):
     return True
 
 
+# Inpainting
+def sd_inpaint(scene):
+    """Post to the API to generate a Stable Diffusion image with inpainting, and then process it"""
+    props = scene.air_props
+
+    # get the prompt if we haven't been given one
+    if props.use_animated_prompts:
+        prompt, negative_prompt = validate_and_process_animated_prompt_text_for_single_frame(scene, scene.frame_current)
+        if not prompt:
+            return False
+    else:
+        prompt = get_full_prompt(scene)
+        negative_prompt = props.negative_prompt_text.strip()
+
+
+    # validate the parameters we will send
+    if not validate_params(scene, prompt):
+        return False
+
+    # generate a new seed, if we want a random one
+    generate_new_random_seed(scene)
+
+    # prepare the output filenames
+    timestamp = int(time.time())
+    before_output_filename_prefix = f"ai-render-{timestamp}-1-before"
+    after_output_filename_prefix = f"ai-render-{timestamp}-2-inpainted"
+    animation_output_filename_prefix = "ai-render-"
+
+    # if we want to use the last SD image, try loading it now
+    if not props.last_generated_image_filename:
+        return handle_error("Couldn't find the last Stable Diffusion image", "last_generated_image_filename")
+    try:
+        img_file = open(props.last_generated_image_filename, 'rb')
+    except:
+        return handle_error("Couldn't load the last Stable Diffusion image. It's probably been deleted or moved. You'll need to restore it or render a new image.", "load_last_generated_image")
+
+    # load mask here
+    if props.inpaint_mask_path == "":
+        return handle_error("Couldn't find the Inpaint Mask File", "inpaint_mask_path")
+    try:
+        mask_file = open(props.inpaint_mask_path, 'rb')
+    except:
+        return handle_error("Couldn't load the uploaded inpaint mask file", "inpaint_mask_path")
+
+    # prepare data for the API request
+    params = {
+        "prompt": prompt,
+        "negative_prompt": negative_prompt,
+        "width": utils.get_output_width(scene),
+        "height": utils.get_output_height(scene),
+        "seed": props.seed,
+        "cfg_scale": props.cfg_scale,
+        "steps": props.steps,
+        "is_full_res" : props.inpaint_full_res,
+        "full_res_padding" : props.inpaint_padding,
+    }
+
+    # get the backend we're using
+    sd_backend = utils.get_active_backend()
+
+    # send to whichever API we're using
+    start_time = time.time()
+    generated_image_file = sd_backend.inpaint(params, img_file, mask_file, after_output_filename_prefix, props)
+
+    # if we didn't get a successful image, stop here (an error will have been handled by the api function)
+    if not generated_image_file:
+        return False
+
+    # autosave the after image, if we should
+    if utils.should_autosave_after_image(props):
+        generated_image_file = save_after_image(scene, after_output_filename_prefix, generated_image_file)
+
+    # store this image filename as the last generated image
+    props.last_generated_image_filename = generated_image_file
+
+    # if we're rendering an animation manually, save the image to the animation output path
+    if props.is_rendering_animation_manually:
+        generated_image_file = save_animation_image(scene, animation_output_filename_prefix, generated_image_file)
+
+    # load the image into our scene
+    try:
+        img = bpy.data.images.load(generated_image_file, check_existing=False)
+    except:
+        return handle_error("Couldn't load the image from Stable Diffusion", "load_sd_image")
+
+    # view the image in the AIR workspace
+    try:
+        utils.view_sd_result_in_air_image_editor(img)
+    except:
+        return handle_error("Couldn't switch the view to the image from Stable Diffusion", "view_sd_image")
+
+    # return success
+    return True
+
+
+# Outpainting
+def sd_outpaint(scene):
+    """Post to the API to generate a Stable Diffusion image with outpainting, and then process it"""
+    props = scene.air_props
+
+    # get the prompt if we haven't been given one
+    if props.use_animated_prompts:
+        prompt, negative_prompt = validate_and_process_animated_prompt_text_for_single_frame(scene, scene.frame_current)
+        if not prompt:
+            return False
+    else:
+        prompt = get_full_prompt(scene)
+        negative_prompt = props.negative_prompt_text.strip()
+
+
+    # validate the parameters we will send
+    if not validate_params(scene, prompt):
+        return False
+
+    # generate a new seed, if we want a random one
+    generate_new_random_seed(scene)
+
+    # prepare the output filenames
+    timestamp = int(time.time())
+    before_output_filename_prefix = f"ai-render-{timestamp}-1-before"
+    after_output_filename_prefix = f"ai-render-{timestamp}-2-outpainted"
+    animation_output_filename_prefix = "ai-render-"
+
+    # if we want to use the last SD image, try loading it now
+    if not props.last_generated_image_filename:
+        return handle_error("Couldn't find the last Stable Diffusion image", "last_generated_image_filename")
+    try:
+        img_file = open(props.last_generated_image_filename, 'rb')
+    except:
+        return handle_error("Couldn't load the last Stable Diffusion image. It's probably been deleted or moved. You'll need to restore it or render a new image.", "load_last_generated_image")
+
+
+    # prepare data for the API request
+    params = {
+        "prompt": prompt,
+        "negative_prompt": negative_prompt,
+        "width": utils.get_output_width(scene),
+        "height": utils.get_output_height(scene),
+        "seed": props.seed,
+        "cfg_scale": props.cfg_scale,
+        "steps": props.steps,
+        "pixels": props.outpaint_pixels_to_expand,
+        "mask_blur": props.outpaint_mask_blur,
+        "directions": [props.outpaint_direction],
+        "noise_q": props.outpaint_noise_q,
+        "color_variation": props.outpaint_color_variation,
+    }
+
+    # get the backend we're using
+    sd_backend = utils.get_active_backend()
+
+    # send to whichever API we're using
+    start_time = time.time()
+    generated_image_file = sd_backend.outpaint(params, img_file, after_output_filename_prefix, props)
+
+    # if we didn't get a successful image, stop here (an error will have been handled by the api function)
+    if not generated_image_file:
+        return False
+
+    # autosave the after image, if we should
+    if utils.should_autosave_after_image(props):
+        generated_image_file = save_after_image(scene, after_output_filename_prefix, generated_image_file)
+
+    # store this image filename as the last generated image
+    props.last_generated_image_filename = generated_image_file
+
+    # if we're rendering an animation manually, save the image to the animation output path
+    if props.is_rendering_animation_manually:
+        generated_image_file = save_animation_image(scene, animation_output_filename_prefix, generated_image_file)
+
+    # load the image into our scene
+    try:
+        img = bpy.data.images.load(generated_image_file, check_existing=False)
+    except:
+        return handle_error("Couldn't load the image from Stable Diffusion", "load_sd_image")
+
+    # view the image in the AIR workspace
+    try:
+        utils.view_sd_result_in_air_image_editor(img)
+    except:
+        return handle_error("Couldn't switch the view to the image from Stable Diffusion", "view_sd_image")
+
+    # return success
+    return True
+
 
 class AIR_OT_enable(bpy.types.Operator):
     "Enable AI Render in this scene"
@@ -970,6 +1155,35 @@ def execute(self, context):
         return {'FINISHED'}
 
 
+class AIR_OT_inpaint_from_last_sd_image(bpy.types.Operator):
+    "Inpaint a new Stable Diffusion image - without re-rendering - using the most recent Stable Diffusion image as the starting point"
+    bl_idname = "ai_render.inpaint_from_last_sd_image"
+    bl_label = "Inpaint Image From Last AI Image"
+
+    def execute(self, context):
+        do_pre_render_setup(context.scene)
+        do_pre_api_setup(context.scene)
+
+        # post to the api (on a different thread, outside the operator)
+        task_queue.add(functools.partial(sd_inpaint, context.scene))
+
+        return {'FINISHED'}
+
+
+class AIR_OT_outpaint_from_last_sd_image(bpy.types.Operator):
+    "Inpaint a new Stable Diffusion image - without re-rendering - using the most recent Stable Diffusion image as the starting point"
+    bl_idname = "ai_render.outpaint_from_last_sd_image"
+    bl_label = "Outpaint Image From Last AI Image"
+
+    def execute(self, context):
+        do_pre_render_setup(context.scene)
+        do_pre_api_setup(context.scene)
+
+        # post to the api (on a different thread, outside the operator)
+        task_queue.add(functools.partial(sd_outpaint, context.scene))
+
+        return {'FINISHED'}
+
 
 classes = [
     AIR_OT_enable,
@@ -989,6 +1203,8 @@ def execute(self, context):
     AIR_OT_automatic1111_load_controlnet_models,
     AIR_OT_automatic1111_load_controlnet_modules,
     AIR_OT_automatic1111_load_controlnet_models_and_modules,
+    AIR_OT_inpaint_from_last_sd_image,
+    AIR_OT_outpaint_from_last_sd_image,
 ]
 
 

properties.py

@@ -39,6 +39,15 @@ def get_available_controlnet_modules(self, context):
         return []
 
 
+def get_outpaint_directions(self, context):
+    return [
+        ("up", "up", ""),
+        ("down", "down", ""),
+        ("left", "left", ""),
+        ("right", "right", ""),
+    ]
+
+
 def ensure_sampler(context):
     # """Ensure that the sampler is set to a valid value"""
     scene = context.scene
@@ -275,6 +284,60 @@ class AIRProperties(bpy.types.PropertyGroup):
         items=get_available_controlnet_modules,
         description="Which ControlNet module (preprocessor) to use (these come with the ControlNet extension)",
     )
+    inpaint_mask_path: bpy.props.StringProperty(
+        name="Inpaint Mask Path",
+        default="",
+        description="Upload Inpaint Mask",
+        subtype="FILE_PATH",
+    )
+    inpaint_full_res: bpy.props.BoolProperty(
+        name="Inpaint at Full Resolution",
+        default=True,
+        description="",
+    )
+    inpaint_padding: bpy.props.IntProperty(
+        name="Inpaint Padding",
+        max=256,
+        min=0,
+        default=32,
+        step=4,
+        description="",
+    )
+    outpaint_direction: bpy.props.EnumProperty(
+        name="Outpaint Direction",
+        items=get_outpaint_directions,
+        description="The image will expand in this direction",
+    )
+    outpaint_pixels_to_expand: bpy.props.IntProperty(
+        name="Outpaint Pixels to Expand",
+        min=8,
+        max=256,
+        step=8,
+        default=8,
+        description="",
+    )
+    outpaint_mask_blur: bpy.props.IntProperty(
+        name="Outpaint Mask Blur",
+        description="this changes how much the inpainting mask is blurred. This helps to avoid sharp edges on the image.",
+        min=0,
+        max=64,
+        step=1,
+        default=0,
+    )
+    outpaint_noise_q: bpy.props.FloatProperty(
+        min=0.0,
+        max=4.0,
+        default=1.0,
+        step=0.01,
+        name="Outpaint Noise Quotient",
+    )
+    outpaint_color_variation: bpy.props.FloatProperty(
+        min=0.0,
+        max=1.0,
+        default=0.05,
+        step=0.01,
+        name="Outpaint Color Variation",
+    )
 
 
 classes = [

sd_backends/automatic1111_api.py

@@ -279,6 +279,14 @@ def supports_reloading_upscaler_models():
     return True
 
 
+def supports_inpainting():
+    return False
+
+
+def supports_outpainting():
+    return False
+
+
 def min_image_size():
     return 128 * 128