Update README.md
Browse files
README.md
CHANGED
@@ -10,17 +10,17 @@ tags:
|
|
10 |
- text-to-video
|
11 |
- video
|
12 |
- video-generation
|
13 |
-
base_model: "Wan-AI/Wan2.1-
|
14 |
pipeline_tag: text-to-video
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
instance_prompt:
|
21 |
---
|
22 |
|
23 |
-
# Wan 14B Cyberpunk Realistic
|
24 |
|
25 |
<Gallery />
|
26 |
|
@@ -35,7 +35,7 @@ It was trained on [Replicate](https://replicate.com/) using AI toolkit: https://
|
|
35 |
|
36 |
## Trigger words
|
37 |
|
38 |
-
You should use `
|
39 |
|
40 |
|
41 |
## Use this LoRA
|
@@ -51,8 +51,9 @@ Replicate has a collection of Wan2.1 models that are optimised for speed and cos
|
|
51 |
import replicate
|
52 |
|
53 |
input = {
|
54 |
-
"prompt": "
|
55 |
-
"
|
|
|
56 |
}
|
57 |
|
58 |
output = replicate.run(
|
@@ -88,7 +89,7 @@ pipe.load_lora_weights("fofr/wan-14b-cyberpunk-realistic")
|
|
88 |
|
89 |
pipe.enable_model_cpu_offload() #for low-vram environments
|
90 |
|
91 |
-
prompt = "
|
92 |
negative_prompt = "Bright tones, overexposed, static, blurred details, subtitles, style, works, paintings, images, static, overall gray, worst quality, low quality, JPEG compression residue, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn faces, deformed, disfigured, misshapen limbs, fused fingers, still picture, messy background, three legs, many people in the background, walking backwards"
|
93 |
|
94 |
output = pipe(
|
@@ -105,7 +106,7 @@ export_to_video(output, "output.mp4", fps=16)
|
|
105 |
|
106 |
## Training details
|
107 |
|
108 |
-
- Steps:
|
109 |
- Learning rate: 0.0001
|
110 |
- LoRA rank: 32
|
111 |
|
|
|
10 |
- text-to-video
|
11 |
- video
|
12 |
- video-generation
|
13 |
+
base_model: "Wan-AI/Wan2.1-I2V-14B-480P-Diffusers"
|
14 |
pipeline_tag: text-to-video
|
15 |
+
widget:
|
16 |
+
- text: >-
|
17 |
+
in the style of CYB77, driving a car very fast through a city at night
|
18 |
+
output:
|
19 |
+
url: https://replicate.delivery/xezq/asl3HD14zJpjCV3WAefO5cMDOqyxKG2ZWIfN1dD9DRby4V2oA/R8_Wan_00001.mp4
|
20 |
+
instance_prompt: CYB77
|
21 |
---
|
22 |
|
23 |
+
# Wan 14B Cyberpunk Realistic (image-to-video)
|
24 |
|
25 |
<Gallery />
|
26 |
|
|
|
35 |
|
36 |
## Trigger words
|
37 |
|
38 |
+
You should use `CYB77` or "In the style ocf CYB77" to trigger the video generation.
|
39 |
|
40 |
|
41 |
## Use this LoRA
|
|
|
51 |
import replicate
|
52 |
|
53 |
input = {
|
54 |
+
"prompt": "CYB77",
|
55 |
+
"image": "https://replicate.delivery/xezq/4BGR8w4ELvJfcqIG4KIF0Kr82JxtfQCM2xVAIUieSeifmTZjC/output_frame.jpg",
|
56 |
+
"lora_url": "https://huggingface.co/fofr/wan-14b-cyberpunk-realistic/resolve/main/wan-14b-i2v-cyberpunk-realistic-lora.safetensors"
|
57 |
}
|
58 |
|
59 |
output = replicate.run(
|
|
|
89 |
|
90 |
pipe.enable_model_cpu_offload() #for low-vram environments
|
91 |
|
92 |
+
prompt = "CYB77"
|
93 |
negative_prompt = "Bright tones, overexposed, static, blurred details, subtitles, style, works, paintings, images, static, overall gray, worst quality, low quality, JPEG compression residue, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn faces, deformed, disfigured, misshapen limbs, fused fingers, still picture, messy background, three legs, many people in the background, walking backwards"
|
94 |
|
95 |
output = pipe(
|
|
|
106 |
|
107 |
## Training details
|
108 |
|
109 |
+
- Steps: 5000
|
110 |
- Learning rate: 0.0001
|
111 |
- LoRA rank: 32
|
112 |
|