Initial import

.gitignore

@@ -0,0 +1,2 @@
+.env
+.flagged

app.py

@@ -0,0 +1,35 @@
+import gradio as gr
+from diffusers import DiffusionPipeline
+import cv2
+import torch
+import numpy as np
+from PIL import Image
+
+model = "shadowlamer/sd-zxspectrum-model-256"
+
+# Borrowed from here: https://stackoverflow.com/a/73667318
+def quantize_to_palette(_image, _palette):
+    x_query = _image.reshape(-1, 3).astype(np.float32)
+    x_index = _palette.astype(np.float32)
+    knn = cv2.ml.KNearest_create()
+    knn.train(x_index, cv2.ml.ROW_SAMPLE, np.arange(len(_palette)))
+    ret, results, neighbours, dist = knn.findNearest(x_query, 1)
+    _quantized_image = np.array([_palette[idx] for idx in neighbours.astype(int)])
+    _quantized_image = _quantized_image.reshape(_image.shape)
+    return Image.fromarray(cv2.cvtColor(np.array(_quantized_image, dtype=np.uint8), cv2.COLOR_BGR2RGB))
+
+
+def generate(prompt, seed):
+    pipe = DiffusionPipeline.from_pretrained(model, safety_checker=None, requires_safety_checker=False)
+    generator = torch.Generator("cpu").manual_seed(int(seed))
+    raw_image = pipe(prompt, height=192, width=256, num_inference_steps=20, generator=generator).images[0]
+    palette = np.array(
+        [[0, 0, 0], [0, 0, 255], [0, 255, 0], [0, 255, 255], [255, 0, 0], [255, 0, 255], [255, 255, 0],
+         [255, 255, 255]])
+    input_image = np.array(raw_image)
+    input_image = input_image[:, :, ::-1].copy()
+    return quantize_to_palette(_image=input_image, _palette=palette)
+
+
+iface = gr.Interface(fn=generate, inputs=["text", "number"],  outputs="image")
+iface.launch()

requireents.txt

@@ -0,0 +1,7 @@
+gradio
+diffusers
+torch
+transformers
+accelerate
+opencv-python
+numpy