ws_01.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "d4e99064-36ef-4c7d-9591-cc93ef206e2b",
   "metadata": {
    "jp-MarkdownHeadingCollapsed": true
   },
   "source": [
    "# VSB,FEI - Generative AI Workshop\n",
    "\n",
    "The aim of the workshop is to get an overview of possibilities in the generative artificial intelligence (GenAI) domain.\n",
    "\n",
    "## 🔎 Do you know any famous models from this area?\n",
    "\n",
    "* We will use [Python](https://www.python.org/) and [Huggingface](https://huggingface.co/).\n",
    "\n",
    "![meme01](https://github.com/rasvob/PopAI-FEI-Workshop/blob/main/images/ai_meme_02.jpg?raw=true)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ed6799b8-1842-42a2-9bea-a1a131a7f1cd",
   "metadata": {},
   "source": [
    "## Let's install the basic libraries first\n",
    "\n",
    "* We will use HuggingFace library for the **Stable Diffusion** model\n",
    "    * https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0\n",
    "    * 🔎 What is a **Text-to-Image** task?\n",
    "\n",
    "* HuggingFace is an open source data science and machine learning platform.\n",
    "    * It acts as a hub for AI experts and enthusiasts—like a GitHub for AI.\n",
    "\n",
    "* You can download pre-trained models from the Hub and use them leveraging simple unified API\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "afd9d18c-466f-417d-89ea-aaabd1d7c78d",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "!pip install transformers diffusers\n",
    "!pip install invisible_watermark transformers accelerate safetensors"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cdff7a0f-a289-403b-8964-c1d46c47c7e8",
   "metadata": {},
   "source": [
    "## Running own Stable Diffusion instance is quite easy\n",
    "* You just need to download the pretrained model and load it into the GPU\n",
    "* There are many different models in the [HuggingFace Models Hub](https://huggingface.co/models)\n",
    "    * 💡 Filter by task, e.g. Text-to-Image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cd7df121-adb8-4aac-8e61-6d7af55b25a7",
   "metadata": {},
   "outputs": [],
   "source": [
    "from diffusers import DiffusionPipeline\n",
    "import torch\n",
    "\n",
    "pipe = DiffusionPipeline.from_pretrained(\"stabilityai/stable-diffusion-xl-base-1.0\", torch_dtype=torch.float16, use_safetensors=True, variant=\"fp16\")\n",
    "pipe.to(\"cuda\")\n",
    "\n",
    "## You can specify the device to use\n",
    "# pipe.to(\"cuda:1\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d7730fba-19f8-4640-8c03-42728217c68f",
   "metadata": {},
   "source": [
    "## Now the model is ready and you can start to use it\n",
    "* 💡 The most important part is so-called **prompt** definition - the same concept as in the ChatGPT"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f79b9fed-8979-404f-a639-7f287bcb4d73",
   "metadata": {},
   "source": [
    "# Let's create our own image using the model\n",
    "\n",
    "#### 💡 TIP: Run the code multiple times if you do not like the result 🙂"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "849f2cb6-9fe3-46ef-becb-bf61767d99e5",
   "metadata": {},
   "outputs": [],
   "source": [
    "prompt = \"A knight riding a majestic lion\"\n",
    "\n",
    "image = pipe(prompt=prompt).images[0]\n",
    "\n",
    "image"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6c383046-c2cf-4fa8-bea8-55dc37cd95cf",
   "metadata": {},
   "source": [
    "## The image can be very easily saved"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "60d11b6a-7e88-4cc2-8cb8-d1f5b881486f",
   "metadata": {},
   "outputs": [],
   "source": [
    "image.save('sd_output.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c4a4ea7c-6e4d-42cb-9627-0fbe519ad7a4",
   "metadata": {},
   "source": [
    "# The most difficult part is to define the prompt\n",
    "* There are several *tips&tricks* how to get maximum out of the model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5354c468-e58a-4f5a-ab4f-380b427154b4",
   "metadata": {},
   "source": [
    "## You can add keywords after the main prompt delimited by commas to be more specific"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2283ae84-a8f7-4d12-9430-01cacdb614ed",
   "metadata": {},
   "outputs": [],
   "source": [
    "prompt = \"A knight riding a majestic lion, cyberpunk, japan city background\"\n",
    "\n",
    "image = pipe(prompt=prompt).images[0]\n",
    "\n",
    "image"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3204a979-8ab7-4026-bfb9-01090dd110e0",
   "metadata": {},
   "source": [
    "## You can put an emphasis on a keyword by adding [] [around]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0e52a009-e8b6-4547-8c46-5fa3676c2539",
   "metadata": {},
   "outputs": [],
   "source": [
    "prompt = \"A knight riding a majestic lion, cyberpunk, japan city background, [black and white]\"\n",
    "\n",
    "image = pipe(prompt=prompt).images[0]\n",
    "\n",
    "image"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1c7e32b6-9415-435c-bada-d8ac288dc6fd",
   "metadata": {},
   "source": [
    "# Very common use-case is to use some specific painter style\n",
    "* You can see results of multiple styles at https://www.urania.ai/top-sd-artists"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ef345a05-ee87-452c-b41f-e28e227cf85b",
   "metadata": {},
   "source": [
    "## Leonid Afremov"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3d405b57-79be-41a7-993a-a7df9d1bebd0",
   "metadata": {},
   "outputs": [],
   "source": [
    "prompt = \"A knight riding a majestic lion, [Leonid Afremov]\"\n",
    "\n",
    "image = pipe(prompt=prompt).images[0]\n",
    "\n",
    "image"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "65a8ac00-c9ab-407a-8d80-81f3a2d87a7a",
   "metadata": {},
   "source": [
    "## Vincent Van Gogh"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0c151b8c-88fa-4156-8569-037e24ad8581",
   "metadata": {},
   "outputs": [],
   "source": [
    "prompt = \"A knight riding a majestic lion, [Vincent Van Gogh], cyberpunk\"\n",
    "\n",
    "image = pipe(prompt=prompt).images[0]\n",
    "\n",
    "image"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3d9f9f1f",
   "metadata": {},
   "source": [
    "![meme02](https://github.com/rasvob/PopAI-FEI-Workshop/blob/main/images/ai_meme_01.jpg?raw=true)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "309116ed-9681-4487-99ea-0c5c465f706b",
   "metadata": {},
   "source": [
    "# Sometimes it is good to specify if you want to have 3D or 2D image, otherwise it is random\n",
    "* You can use **2d, sketch** for 2D\n",
    "* **3d, unreal engine, unity, octane render** for 3D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6470d518-11cf-467d-abd5-7e64e2a77aec",
   "metadata": {},
   "outputs": [],
   "source": [
    "prompt = \"A knight riding a majestic lion, 2d, sketch, [color]\"\n",
    "\n",
    "image = pipe(prompt=prompt).images[0]\n",
    "\n",
    "image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "055f60ec-aab8-4f42-9a19-bd1bfa1749b5",
   "metadata": {},
   "outputs": [],
   "source": [
    "prompt = \"A knight riding a majestic lion, 3d, unreal engine, unity, octane render\"\n",
    "\n",
    "image = pipe(prompt=prompt).images[0]\n",
    "\n",
    "image"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fb221ba2-3414-4103-9a9c-172fbb674c99",
   "metadata": {},
   "source": [
    "# If you want to have some more detailed and vibrant image it is possible to add keywords like:\n",
    "* **hyper realism, detailed digital painting, vibrant colors, artstation**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c225cbf9-bef3-4c83-834f-ae347f9473ec",
   "metadata": {},
   "outputs": [],
   "source": [
    "prompt = \"A knight riding a majestic lion, hyper realism, detailed digital painting, vibrant colors, artstation\"\n",
    "\n",
    "image = pipe(prompt=prompt).images[0]\n",
    "image"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c9a8ef5a-d9b1-4593-babb-14869ba687c7",
   "metadata": {},
   "source": [
    "# You can experiment with anything you can imagine 🙂"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1248ebe1-cb3c-4be5-be74-8949e8a7a106",
   "metadata": {},
   "source": [
    "# If you are interested in this topic I recommend to visit [lexica.art](https://lexica.art/) or [https://www.reddit.com/r/StableDiffusion/](https://www.reddit.com/r/StableDiffusion/) 🙂\n",
    "\n",
    "### 💡 You can try it even using web-based GUI on [HuggingFace Spaces](https://huggingface.co/spaces/stabilityai/stable-diffusion)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}