File size: 9,538 Bytes
b177a48
 
 
 
 
 
 
 
02ff012
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b177a48
02ff012
 
b177a48
 
 
b6dc501
 
 
 
 
 
 
 
b257e01
 
 
b177a48
 
a4b32da
b177a48
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b257e01
b177a48
 
 
b257e01
b6dc501
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b177a48
 
 
 
 
 
 
 
 
 
 
b6dc501
b177a48
 
 
a4b32da
b177a48
 
 
 
 
a4b32da
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
import replicate
from PIL import Image
import requests
import io
import os
import base64

Replicate_MODEl_NAME_MAP = {
    "SDXL": "stability-ai/sdxl:7762fd07cf82c948538e41f63f77d685e02b063e37e496e96eefd46c929f9bdc",
    "SD-v3.0": "stability-ai/stable-diffusion-3",
    "SD-v2.1": "stability-ai/stable-diffusion:ac732df83cea7fff18b8472768c88ad041fa750ff7682a21affe81863cbe77e4",
    "SD-v1.5": "stability-ai/stable-diffusion:b3d14e1cd1f9470bbb0bb68cac48e5f483e5be309551992cc33dc30654a82bb7",
    "SDXL-Lightning": "bytedance/sdxl-lightning-4step:5f24084160c9089501c1b3545d9be3c27883ae2239b6f412990e82d4a6210f8f",
    "Kandinsky-v2.0": "ai-forever/kandinsky-2:3c6374e7a9a17e01afe306a5218cc67de55b19ea536466d6ea2602cfecea40a9",
    "Kandinsky-v2.2": "ai-forever/kandinsky-2.2:ad9d7879fbffa2874e1d909d1d37d9bc682889cc65b31f7bb00d2362619f194a",
    "Proteus-v0.2": "lucataco/proteus-v0.2:06775cd262843edbde5abab958abdbb65a0a6b58ca301c9fd78fa55c775fc019",
    "Playground-v2.0": "playgroundai/playground-v2-1024px-aesthetic:42fe626e41cc811eaf02c94b892774839268ce1994ea778eba97103fe1ef51b8",
    "Playground-v2.5": "playgroundai/playground-v2.5-1024px-aesthetic:a45f82a1382bed5c7aeb861dac7c7d191b0fdf74d8d57c4a0e6ed7d4d0bf7d24",
    "Dreamshaper-xl-turbo": "lucataco/dreamshaper-xl-turbo:0a1710e0187b01a255302738ca0158ff02a22f4638679533e111082f9dd1b615",
    "SDXL-Deepcache": "lucataco/sdxl-deepcache:eaf678fb34006669e9a3c6dd5971e2279bf20ee0adeced464d7b6d95de16dc93",
    "Openjourney-v4": "prompthero/openjourney:ad59ca21177f9e217b9075e7300cf6e14f7e5b4505b87b9689dbd866e9768969",
    "LCM-v1.5": "fofr/latent-consistency-model:683d19dc312f7a9f0428b04429a9ccefd28dbf7785fef083ad5cf991b65f406f",
    "Realvisxl-v3.0": "fofr/realvisxl-v3:33279060bbbb8858700eb2146350a98d96ef334fcf817f37eb05915e1534aa1c", 
    
    "Realvisxl-v2.0": "lucataco/realvisxl-v2.0:7d6a2f9c4754477b12c14ed2a58f89bb85128edcdd581d24ce58b6926029de08",
    "Pixart-Sigma": "cjwbw/pixart-sigma:5a54352c99d9fef467986bc8f3a20205e8712cbd3df1cbae4975d6254c902de1",
    "SSD-1b": "lucataco/ssd-1b:b19e3639452c59ce8295b82aba70a231404cb062f2eb580ea894b31e8ce5bbb6",
    "Open-Dalle-v1.1": "lucataco/open-dalle-v1.1:1c7d4c8dec39c7306df7794b28419078cb9d18b9213ab1c21fdc46a1deca0144",
    "Deepfloyd-IF": "andreasjansson/deepfloyd-if:fb84d659df149f4515c351e394d22222a94144aa1403870c36025c8b28846c8d",

    "Zeroscope-v2-xl": "anotherjesse/zeroscope-v2-xl:9f747673945c62801b13b84701c783929c0ee784e4748ec062204894dda1a351",
    # "Damo-Text-to-Video": "cjwbw/damo-text-to-video:1e205ea73084bd17a0a3b43396e49ba0d6bc2e754e9283b2df49fad2dcf95755",
    "Animate-Diff": "lucataco/animate-diff:beecf59c4aee8d81bf04f0381033dfa10dc16e845b4ae00d281e2fa377e48a9f",
    "OpenSora": "camenduru/open-sora:8099e5722ba3d5f408cd3e696e6df058137056268939337a3fbe3912e86e72ad",
    "LaVie": "cjwbw/lavie:0bca850c4928b6c30052541fa002f24cbb4b677259c461dd041d271ba9d3c517",
    "VideoCrafter2": "lucataco/video-crafter:7757c5775e962c618053e7df4343052a21075676d6234e8ede5fa67c9e43bce0",
    "Stable-Video-Diffusion": "sunfjun/stable-video-diffusion:d68b6e09eedbac7a49e3d8644999d93579c386a083768235cabca88796d70d82",
    "FLUX.1-schnell": "black-forest-labs/flux-schnell",
    "FLUX.1-pro": "black-forest-labs/flux-pro",
    "FLUX.1-dev": "black-forest-labs/flux-dev",
    }


class ReplicateModel():
    def __init__(self, model_name, model_type):
        self.model_name = model_name
        self.model_type = model_type
    
    def __call__(self, *args, **kwargs):
        if self.model_type == "text2image":
            assert "prompt" in kwargs, "prompt is required for text2image model"
            output  = replicate.run(
                                    f"{Replicate_MODEl_NAME_MAP[self.model_name]}",
                                    input={
                                            "width": 512,
                                            "height": 512,
                                            "prompt": kwargs["prompt"]
                                        },
                                )
            if 'Openjourney' in self.model_name:
                for item in output:
                    result_url = item
                    break
            elif isinstance(output, list):
                result_url = output[0]
            else:
                result_url = output
            print(self.model_name, result_url)
            response = requests.get(result_url)
            result = Image.open(io.BytesIO(response.content))
            return result

        elif self.model_type == "text2video":
            assert "prompt" in kwargs, "prompt is required for text2image model"
            if self.model_name == "Zeroscope-v2-xl":
                input = {
                            "fps": 24,
                            "width": 512,
                            "height": 512,
                            "prompt": kwargs["prompt"],
                            "guidance_scale": 17.5,
                            # "negative_prompt": "very blue, dust, noisy, washed out, ugly, distorted, broken",
                            "num_frames": 48,
                        }
            elif self.model_name == "Damo-Text-to-Video":
                input={
                    "fps": 8,
                    "prompt": kwargs["prompt"],
                    "num_frames": 16,
                    "num_inference_steps": 50
                }
            elif self.model_name == "Animate-Diff":
                input={
                    "path": "toonyou_beta3.safetensors",
                    "seed": 255224557,
                    "steps": 25,
                    "prompt": kwargs["prompt"],
                    "n_prompt": "badhandv4, easynegative, ng_deepnegative_v1_75t, verybadimagenegative_v1.3, bad-artist, bad_prompt_version2-neg, teeth",
                    "motion_module": "mm_sd_v14",
                    "guidance_scale": 7.5
                }
            elif self.model_name == "OpenSora":
                input={
                    "seed": 1234,
                    "prompt": kwargs["prompt"],
                }
            elif self.model_name == "LaVie":
                input={
                    "width": 512,
                    "height": 512,
                    "prompt": kwargs["prompt"],
                    "quality": 9,
                    "video_fps": 8,
                    "interpolation": False,
                    "sample_method": "ddpm",
                    "guidance_scale": 7,
                    "super_resolution": False,
                    "num_inference_steps": 50
                }
            elif self.model_name == "VideoCrafter2":
                input={
                    "fps": 24,
                    "seed": 64045,
                    "steps": 40,
                    "width": 512,
                    "height": 512,
                    "prompt": kwargs["prompt"],
                }
            elif self.model_name == "Stable-Video-Diffusion":
                text2image_name = "SD-v2.1"
                output = replicate.run(
                                        f"{Replicate_MODEl_NAME_MAP[text2image_name]}",
                                        input={
                                                "width": 512,
                                                "height": 512,
                                                "prompt": kwargs["prompt"]
                                            },
                                    )
                if isinstance(output, list):
                    image_url = output[0]
                else:
                    image_url = output
                print(image_url)

                input={
                    "cond_aug": 0.02,
                    "decoding_t": 14,
                    "input_image": "{}".format(image_url),
                    "video_length": "14_frames_with_svd",
                    "sizing_strategy": "maintain_aspect_ratio",
                    "motion_bucket_id": 127,
                    "frames_per_second": 6
                }

            output  = replicate.run(
                                    f"{Replicate_MODEl_NAME_MAP[self.model_name]}",
                                    input=input,
                                )
            if isinstance(output, list):
                result_url = output[0]
            else:
                result_url = output
            print(self.model_name)
            print(result_url)
            # response = requests.get(result_url)
            # result = Image.open(io.BytesIO(response.content))

        #     for event in handler.iter_events(with_logs=True):
        #         if isinstance(event, fal_client.InProgress):
        #             print('Request in progress')
        #             print(event.logs)

        #     result = handler.get()
        #     print("result video: ====")
        #     print(result)
        #     result_url = result['video']['url']
        #     return result_url
            return result_url
        else:
            raise ValueError("model_type must be text2image or image2image")


def load_replicate_model(model_name, model_type):
    return ReplicateModel(model_name, model_type)


if __name__ == "__main__":
    model_name = 'replicate_zeroscope-v2-xl_text2video'
    model_source, model_name, model_type = model_name.split("_")
    pipe = load_replicate_model(model_name, model_type)
    prompt = "Clown fish swimming in a coral reef, beautiful, 8k, perfect, award winning, national geographic"
    result = pipe(prompt=prompt)