Skip to content

Commit

Permalink
Add tutorial code
Browse files Browse the repository at this point in the history
  • Loading branch information
martin-martin committed Nov 14, 2022
1 parent 87758dd commit fbcd91f
Show file tree
Hide file tree
Showing 7 changed files with 132 additions and 0 deletions.
35 changes: 35 additions & 0 deletions openai-dalle/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,35 @@
# Generate Images With DALL·E 2 and the OpenAI API

Learn to use the OpenAI Python library to create images with DALL·E, a state-of-the-art latent diffusion model. In the associated tutorial on [generating images with DALL·E 2 and the OpenAI API](https://realpython.com/generate-images-with-dalle-openai-api/), you explore image creation and generating image variations. You learn to interact with DALL·E using API calls, and incorporate this functionality into your Python scripts.

## Setup

Create and activate a virtual environment, then install the `openai` package:

```console
$ python --version
Python 3.11.0
$ python -m venv venv
$ source venv/bin/activate
(venv) $ python -m pip install openai
```

You need to be on Python 3.7.1 or higher.

## Create Images and Image Variations

Follow the instructions in [the tutorial](https://realpython.com/generate-images-with-dalle-openai-api/) to create images from text prompts, create image variations, and convert a Base64 JSON response to a PNG image file.

You can find the code for each of these steps in dedicated scripts:

- `create.py`: Create an image from a text prompt and save the image data to a file.
- `convert.py`: Convert a Base64 encoded PNG image delivered in a JSON response to a PNG image file.
- `vary.py`: Read Base64 encoded image data and make an API request to receive variations of that image.

The tutorial walks you through each of these scripts and their functionality and output in more detail.

## Edit Images (Inpainting and Outpainting)

The OpenAI Image API also allows you to [edit parts of an image](https://beta.openai.com/docs/guides/images/edits) using text prompts. For this, you need to create a mask with transparent image data in the area where you want to edit the image.

You can run `edit.py` to give this functionality a try.
17 changes: 17 additions & 0 deletions openai-dalle/convert.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,17 @@
import json
from base64 import b64decode
from pathlib import Path

JSON_FILE_NAME = "An ec-1667994848"
IMAGE_DIR = Path.cwd() / "images" / JSON_FILE_NAME
DATA_DIR = Path.cwd() / "responses"

IMAGE_DIR.mkdir(parents=True, exist_ok=True)

with open(DATA_DIR / f"{JSON_FILE_NAME}.json", "r") as f:
response = json.load(f)

for index, image_dict in enumerate(response["data"]):
image_data = b64decode(image_dict["b64_json"])
with open(IMAGE_DIR / f"{JSON_FILE_NAME}-{index}.png", "wb") as f:
f.write(image_data)
24 changes: 24 additions & 0 deletions openai-dalle/create.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,24 @@
import json
import os
from pathlib import Path

import openai

PROMPT = "An eco-friendly computer from the 90s in the style of vaporwave"
DATA_DIR = Path.cwd() / "responses"

DATA_DIR.mkdir(exist_ok=True)

openai.api_key = os.getenv("OPENAI_API_KEY")

response = openai.Image.create(
prompt=PROMPT,
n=1,
size="256x256",
response_format="b64_json",
)

file_name = DATA_DIR / f"{PROMPT[:5]}-{response['created']}.json"

with open(file_name, mode="w") as file_path:
json.dump(response, file_path)
29 changes: 29 additions & 0 deletions openai-dalle/edit.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,29 @@
import json
import os
from pathlib import Path

import openai

SOURCE_PATH = Path.cwd() / "images" / "An ec-1667994848"
DESTINATION_PATH = Path.cwd() / "responses"
PROMPT = "A 90s vaporwave computer showing Rick Astley on the screen"

SOURCE_PATH.mkdir(parents=True, exist_ok=True)
DESTINATION_PATH.mkdir(parents=True, exist_ok=True)

openai.api_key = os.getenv("OPENAI_API_KEY")

response = openai.Image.create_edit(
image=open(SOURCE_PATH / "computer.png", "rb"),
mask=open(SOURCE_PATH / "mask.png", "rb"),
prompt=PROMPT,
n=1,
size="256x256",
response_format="b64_json",
)

with open(
DESTINATION_PATH / f"edit-{PROMPT[:5]}-{response['created']}.json",
mode="w",
) as fp:
json.dump(response, fp)
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file added openai-dalle/images/An ec-1667994848/mask.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
27 changes: 27 additions & 0 deletions openai-dalle/vary.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,27 @@
import json
import os
from base64 import b64decode
from pathlib import Path

import openai

SOURCE_FILE = "A com-1667994848.json"
DATA_DIR = Path.cwd() / "responses"

openai.api_key = os.getenv("OPENAI_API_KEY")

with open(DATA_DIR / SOURCE_FILE, "r") as json_file:
saved_response = json.load(json_file)
image_data = b64decode(saved_response["data"][0]["b64_json"])

response = openai.Image.create_variation(
image=image_data,
n=3,
size="256x256",
response_format="b64_json",
)

new_file_name = f"vary-{SOURCE_FILE[:4]}-{response['created']}.json"

with open(DATA_DIR / new_file_name, mode="w") as new_json_file:
json.dump(response, new_json_file)

0 comments on commit fbcd91f

Please sign in to comment.