nodes.py 16.5 KB
Newer Older
comfyanonymous's avatar
comfyanonymous committed
1
2
3
4
5
import torch

import os
import sys
import json
6
import hashlib
comfyanonymous's avatar
comfyanonymous committed
7
import copy
comfyanonymous's avatar
comfyanonymous committed
8
9
10
11
12
13
14
15
16
17
18
19

from PIL import Image
from PIL.PngImagePlugin import PngInfo
import numpy as np

sys.path.append(os.path.join(sys.path[0], "comfy"))


import comfy.samplers
import comfy.sd

supported_ckpt_extensions = ['.ckpt']
comfyanonymous's avatar
comfyanonymous committed
20
supported_pt_extensions = ['.ckpt', '.pt']
comfyanonymous's avatar
comfyanonymous committed
21
22
23
try:
    import safetensors.torch
    supported_ckpt_extensions += ['.safetensors']
comfyanonymous's avatar
comfyanonymous committed
24
    supported_pt_extensions += ['.safetensors']
comfyanonymous's avatar
comfyanonymous committed
25
26
27
28
29
30
31
32
33
except:
    print("Could not import safetensors, safetensors support disabled.")

def filter_files_extensions(files, extensions):
    return sorted(list(filter(lambda a: os.path.splitext(a)[-1].lower() in extensions, files)))

class CLIPTextEncode:
    @classmethod
    def INPUT_TYPES(s):
34
        return {"required": {"text": ("STRING", {"multiline": True}), "clip": ("CLIP", )}}
comfyanonymous's avatar
comfyanonymous committed
35
36
37
    RETURN_TYPES = ("CONDITIONING",)
    FUNCTION = "encode"

38
39
    CATEGORY = "conditioning"

comfyanonymous's avatar
comfyanonymous committed
40
    def encode(self, clip, text):
comfyanonymous's avatar
comfyanonymous committed
41
42
43
44
45
46
47
48
49
        return ([[clip.encode(text), {}]], )

class ConditioningCombine:
    @classmethod
    def INPUT_TYPES(s):
        return {"required": {"conditioning_1": ("CONDITIONING", ), "conditioning_2": ("CONDITIONING", )}}
    RETURN_TYPES = ("CONDITIONING",)
    FUNCTION = "combine"

50
51
    CATEGORY = "conditioning"

comfyanonymous's avatar
comfyanonymous committed
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
    def combine(self, conditioning_1, conditioning_2):
        return (conditioning_1 + conditioning_2, )

class ConditioningSetArea:
    @classmethod
    def INPUT_TYPES(s):
        return {"required": {"conditioning": ("CONDITIONING", ),
                              "width": ("INT", {"default": 64, "min": 64, "max": 4096, "step": 64}),
                              "height": ("INT", {"default": 64, "min": 64, "max": 4096, "step": 64}),
                              "x": ("INT", {"default": 0, "min": 0, "max": 4096, "step": 64}),
                              "y": ("INT", {"default": 0, "min": 0, "max": 4096, "step": 64}),
                              "strength": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
                             }}
    RETURN_TYPES = ("CONDITIONING",)
    FUNCTION = "append"

68
69
    CATEGORY = "conditioning"

comfyanonymous's avatar
comfyanonymous committed
70
71
72
73
74
75
76
77
    def append(self, conditioning, width, height, x, y, strength, min_sigma=0.0, max_sigma=99.0):
        c = copy.deepcopy(conditioning)
        for t in c:
            t[1]['area'] = (height // 8, width // 8, y // 8, x // 8)
            t[1]['strength'] = strength
            t[1]['min_sigma'] = min_sigma
            t[1]['max_sigma'] = max_sigma
        return (c, )
comfyanonymous's avatar
comfyanonymous committed
78
79
80
81
82
83
84
85
86
87
88

class VAEDecode:
    def __init__(self, device="cpu"):
        self.device = device

    @classmethod
    def INPUT_TYPES(s):
        return {"required": { "samples": ("LATENT", ), "vae": ("VAE", )}}
    RETURN_TYPES = ("IMAGE",)
    FUNCTION = "decode"

89
90
    CATEGORY = "latent"

comfyanonymous's avatar
comfyanonymous committed
91
92
93
94
95
96
97
98
99
100
101
102
103
    def decode(self, vae, samples):
        return (vae.decode(samples), )

class VAEEncode:
    def __init__(self, device="cpu"):
        self.device = device

    @classmethod
    def INPUT_TYPES(s):
        return {"required": { "pixels": ("IMAGE", ), "vae": ("VAE", )}}
    RETURN_TYPES = ("LATENT",)
    FUNCTION = "encode"

104
105
    CATEGORY = "latent"

comfyanonymous's avatar
comfyanonymous committed
106
    def encode(self, vae, pixels):
107
108
109
110
        x = (pixels.shape[1] // 64) * 64
        y = (pixels.shape[2] // 64) * 64
        if pixels.shape[1] != x or pixels.shape[2] != y:
            pixels = pixels[:,:x,:y,:]
comfyanonymous's avatar
comfyanonymous committed
111
112
113
114
115
116
117
118
119
120
121
122
123
124
        return (vae.encode(pixels), )

class CheckpointLoader:
    models_dir = os.path.join(os.path.dirname(os.path.realpath(__file__)), "models")
    config_dir = os.path.join(models_dir, "configs")
    ckpt_dir = os.path.join(models_dir, "checkpoints")

    @classmethod
    def INPUT_TYPES(s):
        return {"required": { "config_name": (filter_files_extensions(os.listdir(s.config_dir), '.yaml'), ),
                              "ckpt_name": (filter_files_extensions(os.listdir(s.ckpt_dir), supported_ckpt_extensions), )}}
    RETURN_TYPES = ("MODEL", "CLIP", "VAE")
    FUNCTION = "load_checkpoint"

125
126
    CATEGORY = "loaders"

comfyanonymous's avatar
comfyanonymous committed
127
128
129
    def load_checkpoint(self, config_name, ckpt_name, output_vae=True, output_clip=True):
        config_path = os.path.join(self.config_dir, config_name)
        ckpt_path = os.path.join(self.ckpt_dir, ckpt_name)
130
131
        embedding_directory = os.path.join(self.models_dir, "embeddings")
        return comfy.sd.load_checkpoint(config_path, ckpt_path, output_vae=True, output_clip=True, embedding_directory=embedding_directory)
comfyanonymous's avatar
comfyanonymous committed
132
133
134
135
136
137

class VAELoader:
    models_dir = os.path.join(os.path.dirname(os.path.realpath(__file__)), "models")
    vae_dir = os.path.join(models_dir, "vae")
    @classmethod
    def INPUT_TYPES(s):
comfyanonymous's avatar
comfyanonymous committed
138
        return {"required": { "vae_name": (filter_files_extensions(os.listdir(s.vae_dir), supported_pt_extensions), )}}
comfyanonymous's avatar
comfyanonymous committed
139
140
141
    RETURN_TYPES = ("VAE",)
    FUNCTION = "load_vae"

142
143
    CATEGORY = "loaders"

comfyanonymous's avatar
comfyanonymous committed
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
    #TODO: scale factor?
    def load_vae(self, vae_name):
        vae_path = os.path.join(self.vae_dir, vae_name)
        vae = comfy.sd.VAE(ckpt_path=vae_path)
        return (vae,)

class EmptyLatentImage:
    def __init__(self, device="cpu"):
        self.device = device

    @classmethod
    def INPUT_TYPES(s):
        return {"required": { "width": ("INT", {"default": 512, "min": 64, "max": 4096, "step": 64}),
                              "height": ("INT", {"default": 512, "min": 64, "max": 4096, "step": 64}),
                              "batch_size": ("INT", {"default": 1, "min": 1, "max": 64})}}
    RETURN_TYPES = ("LATENT",)
    FUNCTION = "generate"

162
163
    CATEGORY = "latent"

comfyanonymous's avatar
comfyanonymous committed
164
165
166
167
168
169
    def generate(self, width, height, batch_size=1):
        latent = torch.zeros([batch_size, 4, height // 8, width // 8])
        return (latent, )

class LatentUpscale:
    upscale_methods = ["nearest-exact", "bilinear", "area"]
170
    crop_methods = ["disabled", "center"]
comfyanonymous's avatar
comfyanonymous committed
171
172
173
174
175

    @classmethod
    def INPUT_TYPES(s):
        return {"required": { "samples": ("LATENT",), "upscale_method": (s.upscale_methods,),
                              "width": ("INT", {"default": 512, "min": 64, "max": 4096, "step": 64}),
176
177
                              "height": ("INT", {"default": 512, "min": 64, "max": 4096, "step": 64}),
                              "crop": (s.crop_methods,)}}
comfyanonymous's avatar
comfyanonymous committed
178
179
180
    RETURN_TYPES = ("LATENT",)
    FUNCTION = "upscale"

181
182
    CATEGORY = "latent"

183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
    def upscale(self, samples, upscale_method, width, height, crop):
        if crop == "center":
            old_width = samples.shape[3]
            old_height = samples.shape[2]
            old_aspect = old_width / old_height
            new_aspect = width / height
            x = 0
            y = 0
            if old_aspect > new_aspect:
                x = round((old_width - old_width * (new_aspect / old_aspect)) / 2)
            elif old_aspect < new_aspect:
                y = round((old_height - old_height * (old_aspect / new_aspect)) / 2)
            s = samples[:,:,y:old_height-y,x:old_width-x]
        else:
            s = samples
        s = torch.nn.functional.interpolate(s, size=(height // 8, width // 8), mode=upscale_method)
comfyanonymous's avatar
comfyanonymous committed
199
200
        return (s,)

comfyanonymous's avatar
comfyanonymous committed
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
class LatentRotate:
    @classmethod
    def INPUT_TYPES(s):
        return {"required": { "samples": ("LATENT",),
                              "rotation": (["none", "90 degrees", "180 degrees", "270 degrees"],),
                              }}
    RETURN_TYPES = ("LATENT",)
    FUNCTION = "rotate"

    CATEGORY = "latent"

    def rotate(self, samples, rotation):
        rotate_by = 0
        if rotation.startswith("90"):
            rotate_by = 1
        elif rotation.startswith("180"):
            rotate_by = 2
        elif rotation.startswith("270"):
            rotate_by = 3

        s = torch.rot90(samples, k=rotate_by, dims=[3, 2])
        return (s,)
comfyanonymous's avatar
comfyanonymous committed
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243

class LatentFlip:
    @classmethod
    def INPUT_TYPES(s):
        return {"required": { "samples": ("LATENT",),
                              "flip_method": (["x-axis: vertically", "y-axis: horizontally"],),
                              }}
    RETURN_TYPES = ("LATENT",)
    FUNCTION = "flip"

    CATEGORY = "latent"

    def flip(self, samples, flip_method):
        if flip_method.startswith("x"):
            s = torch.flip(samples, dims=[2])
        elif flip_method.startswith("y"):
            s = torch.flip(samples, dims=[3])
        else:
            s = samples

        return (s,)
comfyanonymous's avatar
comfyanonymous committed
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264

class LatentComposite:
    @classmethod
    def INPUT_TYPES(s):
        return {"required": { "samples_to": ("LATENT",),
                              "samples_from": ("LATENT",),
                              "x": ("INT", {"default": 0, "min": 0, "max": 4096, "step": 8}),
                              "y": ("INT", {"default": 0, "min": 0, "max": 4096, "step": 8}),
                              }}
    RETURN_TYPES = ("LATENT",)
    FUNCTION = "composite"

    CATEGORY = "latent"

    def composite(self, samples_to, samples_from, x, y, composite_method="normal"):
        x =  x // 8
        y = y // 8
        s = samples_to.clone()
        s[:,:,y:y+samples_from.shape[2],x:x+samples_from.shape[3]] = samples_from[:,:,:samples_to.shape[2] - y, :samples_to.shape[3] - x]
        return (s,)

comfyanonymous's avatar
comfyanonymous committed
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
def common_ksampler(device, model, seed, steps, cfg, sampler_name, scheduler, positive, negative, latent_image, denoise=1.0, disable_noise=False, start_step=None, last_step=None, force_full_denoise=False):
    if disable_noise:
        noise = torch.zeros(latent_image.size(), dtype=latent_image.dtype, layout=latent_image.layout, device="cpu")
    else:
        noise = torch.randn(latent_image.size(), dtype=latent_image.dtype, layout=latent_image.layout, generator=torch.manual_seed(seed), device="cpu")

    model = model.to(device)
    noise = noise.to(device)
    latent_image = latent_image.to(device)

    positive_copy = []
    negative_copy = []

    for p in positive:
        t = p[0]
        if t.shape[0] < noise.shape[0]:
            t = torch.cat([t] * noise.shape[0])
        t = t.to(device)
        positive_copy += [[t] + p[1:]]
    for n in negative:
        t = n[0]
        if t.shape[0] < noise.shape[0]:
            t = torch.cat([t] * noise.shape[0])
        t = t.to(device)
        negative_copy += [[t] + n[1:]]

    if sampler_name in comfy.samplers.KSampler.SAMPLERS:
        sampler = comfy.samplers.KSampler(model, steps=steps, device=device, sampler=sampler_name, scheduler=scheduler, denoise=denoise)
    else:
        #other samplers
        pass

    samples = sampler.sample(noise, positive_copy, negative_copy, cfg=cfg, latent_image=latent_image, start_step=start_step, last_step=last_step, force_full_denoise=force_full_denoise)
    samples = samples.cpu()
    model = model.cpu()
    return (samples, )

comfyanonymous's avatar
comfyanonymous committed
302
303
304
305
306
307
class KSampler:
    def __init__(self, device="cuda"):
        self.device = device

    @classmethod
    def INPUT_TYPES(s):
comfyanonymous's avatar
comfyanonymous committed
308
        return {"required":
comfyanonymous's avatar
comfyanonymous committed
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
                    {"model": ("MODEL",),
                    "seed": ("INT", {"default": 0, "min": 0, "max": 0xffffffffffffffff}),
                    "steps": ("INT", {"default": 20, "min": 1, "max": 10000}),
                    "cfg": ("FLOAT", {"default": 8.0, "min": 0.0, "max": 100.0}),
                    "sampler_name": (comfy.samplers.KSampler.SAMPLERS, ),
                    "scheduler": (comfy.samplers.KSampler.SCHEDULERS, ),
                    "positive": ("CONDITIONING", ),
                    "negative": ("CONDITIONING", ),
                    "latent_image": ("LATENT", ),
                    "denoise": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01}),
                    }}

    RETURN_TYPES = ("LATENT",)
    FUNCTION = "sample"

324
325
    CATEGORY = "sampling"

comfyanonymous's avatar
comfyanonymous committed
326
    def sample(self, model, seed, steps, cfg, sampler_name, scheduler, positive, negative, latent_image, denoise=1.0):
comfyanonymous's avatar
comfyanonymous committed
327
        return common_ksampler(self.device, model, seed, steps, cfg, sampler_name, scheduler, positive, negative, latent_image, denoise=denoise)
comfyanonymous's avatar
comfyanonymous committed
328

comfyanonymous's avatar
comfyanonymous committed
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
class KSamplerAdvanced:
    def __init__(self, device="cuda"):
        self.device = device

    @classmethod
    def INPUT_TYPES(s):
        return {"required":
                    {"model": ("MODEL",),
                    "add_noise": (["enable", "disable"], ),
                    "noise_seed": ("INT", {"default": 0, "min": 0, "max": 0xffffffffffffffff}),
                    "steps": ("INT", {"default": 20, "min": 1, "max": 10000}),
                    "cfg": ("FLOAT", {"default": 8.0, "min": 0.0, "max": 100.0}),
                    "sampler_name": (comfy.samplers.KSampler.SAMPLERS, ),
                    "scheduler": (comfy.samplers.KSampler.SCHEDULERS, ),
                    "positive": ("CONDITIONING", ),
                    "negative": ("CONDITIONING", ),
                    "latent_image": ("LATENT", ),
                    "start_at_step": ("INT", {"default": 0, "min": 0, "max": 10000}),
                    "end_at_step": ("INT", {"default": 10000, "min": 0, "max": 10000}),
                    "return_with_leftover_noise": (["disable", "enable"], ),
                    }}

    RETURN_TYPES = ("LATENT",)
    FUNCTION = "sample"

    CATEGORY = "sampling"
comfyanonymous's avatar
comfyanonymous committed
355

comfyanonymous's avatar
comfyanonymous committed
356
357
358
359
360
361
362
363
    def sample(self, model, add_noise, noise_seed, steps, cfg, sampler_name, scheduler, positive, negative, latent_image, start_at_step, end_at_step, return_with_leftover_noise, denoise=1.0):
        force_full_denoise = True
        if return_with_leftover_noise == "enable":
            force_full_denoise = False
        disable_noise = False
        if add_noise == "disable":
            disable_noise = True
        return common_ksampler(self.device, model, noise_seed, steps, cfg, sampler_name, scheduler, positive, negative, latent_image, denoise=denoise, disable_noise=disable_noise, start_step=start_at_step, last_step=end_at_step, force_full_denoise=force_full_denoise)
comfyanonymous's avatar
comfyanonymous committed
364
365
366
367
368
369
370
371

class SaveImage:
    def __init__(self):
        self.output_dir = os.path.join(os.path.dirname(os.path.realpath(__file__)), "output")

    @classmethod
    def INPUT_TYPES(s):
        return {"required": 
372
373
                    {"images": ("IMAGE", ),
                     "filename_prefix": ("STRING", {"default": "ComfyUI"})},
comfyanonymous's avatar
comfyanonymous committed
374
375
376
377
378
379
380
381
                "hidden": {"prompt": "PROMPT", "extra_pnginfo": "EXTRA_PNGINFO"},
                }

    RETURN_TYPES = ()
    FUNCTION = "save_images"

    OUTPUT_NODE = True

382
383
    CATEGORY = "image"

384
385
386
387
388
389
390
391
392
393
394
395
396
    def save_images(self, images, filename_prefix="ComfyUI", prompt=None, extra_pnginfo=None):
        def map_filename(filename):
            prefix_len = len(filename_prefix)
            prefix = filename[:prefix_len + 1]
            try:
                digits = int(filename[prefix_len + 1:].split('_')[0])
            except:
                digits = 0
            return (digits, prefix)
        try:
            counter = max(filter(lambda a: a[1][:-1] == filename_prefix and a[1][-1] == "_", map(map_filename, os.listdir(self.output_dir))))[0] + 1
        except ValueError:
            counter = 1
comfyanonymous's avatar
comfyanonymous committed
397
398
399
400
401
402
403
404
405
        for image in images:
            i = 255. * image.cpu().numpy()
            img = Image.fromarray(i.astype(np.uint8))
            metadata = PngInfo()
            if prompt is not None:
                metadata.add_text("prompt", json.dumps(prompt))
            if extra_pnginfo is not None:
                for x in extra_pnginfo:
                    metadata.add_text(x, json.dumps(extra_pnginfo[x]))
406
407
            img.save(f"output/{filename_prefix}_{counter:05}_.png", pnginfo=metadata, optimize=True)
            counter += 1
comfyanonymous's avatar
comfyanonymous committed
408

409
410
411
412
413
414
415
class LoadImage:
    input_dir = os.path.join(os.path.dirname(os.path.realpath(__file__)), "input")
    @classmethod
    def INPUT_TYPES(s):
        return {"required":
                    {"image": (os.listdir(s.input_dir), )},
                }
416
417

    CATEGORY = "image"
418
419
420
421
422
423
424
425
426
427

    RETURN_TYPES = ("IMAGE",)
    FUNCTION = "load_image"
    def load_image(self, image):
        image_path = os.path.join(self.input_dir, image)
        image = Image.open(image_path).convert("RGB")
        image = np.array(image).astype(np.float32) / 255.0
        image = torch.from_numpy(image[None])[None,]
        return image

428
429
430
431
432
433
434
435
    @classmethod
    def IS_CHANGED(s, image):
        image_path = os.path.join(s.input_dir, image)
        m = hashlib.sha256()
        with open(image_path, 'rb') as f:
            m.update(f.read())
        return m.digest().hex()

436

comfyanonymous's avatar
comfyanonymous committed
437
438
439
440
441
442
443
444
445
446
447

NODE_CLASS_MAPPINGS = {
    "KSampler": KSampler,
    "CheckpointLoader": CheckpointLoader,
    "CLIPTextEncode": CLIPTextEncode,
    "VAEDecode": VAEDecode,
    "VAEEncode": VAEEncode,
    "VAELoader": VAELoader,
    "EmptyLatentImage": EmptyLatentImage,
    "LatentUpscale": LatentUpscale,
    "SaveImage": SaveImage,
comfyanonymous's avatar
comfyanonymous committed
448
449
450
    "LoadImage": LoadImage,
    "ConditioningCombine": ConditioningCombine,
    "ConditioningSetArea": ConditioningSetArea,
comfyanonymous's avatar
comfyanonymous committed
451
    "KSamplerAdvanced": KSamplerAdvanced,
comfyanonymous's avatar
comfyanonymous committed
452
    "LatentComposite": LatentComposite,
comfyanonymous's avatar
comfyanonymous committed
453
    "LatentRotate": LatentRotate,
comfyanonymous's avatar
comfyanonymous committed
454
    "LatentFlip": LatentFlip,
comfyanonymous's avatar
comfyanonymous committed
455
456
457
}