Hello Everyone!
This is the 2nd post of a series for my Development Blog, if you want to check my previous post click here: DevBlog #1 Autogenerated descriptions using OpenAI
Introduction
In the previous post I used Python, AWS Lambda and OpenAI API to generate descriptions for TTRPG Terrains Table and store those descriptions in a DynamoDB Table, now we are going to use dalle-2 to generate images using the description as a prompt and store the iamges in AWS S3.
Adding generate image function
For this function we are going to use Image.create function from openai, by default it will return a url that shows the image, but for this case we are going to retrieve the image in base64 so that way we can store it directly in AWS S3.
import os
import random
import boto3
import openai
terrains_table = [#code]
time_table = [#code]
def roll_dice(number):
# code
def get_terrain_by_time(terrain_id, time_id):
# code
def generate_terrain_description(terrain, time):
# code
def generate_terrain_image(description):
openai.api_key = os.getenv("OPENAI_API_KEY")
response = openai.Image.create(
prompt=description,
n=1,
size="512x512",
response_format="b64_json",
)
return response["data"][0]["b64_json"]
def create_terrain(terrain_id, time_id, name, time, description):
# code
def lambda_handler(event, context):
terrain_id = roll_dice(6)
time_id = roll_dice(2)
terrain = terrains_table[terrain_id -1]
time = time_table[time_id -1]
item = get_terrain_by_time(terrain_id, time_id)
if not item:
description = generate_terrain_description(terrain, time)
imageBase64 = generate_terrain_image(description)
item = create_terrain(terrain_id, time_id, terrain, time, description)
return item
At this point we just have the base64 value of the image stored in the imageBase64 variable, now we are going to save the image in AWS S3.
Creating the S3 Bucket and Environment variable
For this we just need to create a simple S3 bucket, no need to specify any properties or permissions as this is just for uploading the images.
Then create a new environment variable named BUCKET and assign the name of your S3 Bucket, as it will be used in the next function.
Also, don't forget to add S3 permissions to your Lambda execution IAM Role, so that way your function will have the required permissions for uploading images.
Adding upload image to s3 function
For this function we are going to specofy the value of the Object Body and assign the value from imageBase64 variable.
import os
import random
import boto3
import openai
import base64
terrains_table = [#code]
time_table = [#code]
def roll_dice(number):
# code
def get_terrain_by_time(terrain_id, time_id):
# code
def generate_terrain_description(terrain, time):
# code
def generate_terrain_image(description):
# code
def upload_base64_to_s3(bucket, key, base_64_str):
s3 = boto3.resource('s3')
s3.Object(bucket, key).put(Body=base64.b64decode(base_64_str))
return f'https://{bucket}.s3.amazonaws.com/{key}'
def create_terrain(terrain_id, time_id, name, time, description):
# code
def lambda_handler(event, context):
terrain_id = roll_dice(6)
time_id = roll_dice(2)
terrain = terrains_table[terrain_id -1]
time = time_table[time_id -1]
item = get_terrain_by_time(terrain_id, time_id)
if not item:
description = generate_terrain_description(terrain, time)
imageBase64 = generate_terrain_image(description)
item = create_terrain(terrain_id, time_id, terrain, time, description)
bucket = os.getenv("BUCKET")
key = f"terrains/{terrain}-{time}.png"
image_url = upload_base64_to_s3(bucket, key, imageBase64)
item = create_terrain(terrain_id, time_id, terrain, time, description, image_url)
return item
Note that we've added a the base64 import and a new property in the create_terrain function, so we need to update that function as well.
Updating create terrain function
import os
import random
import boto3
import openai
import base64
terrains_table = [#code]
time_table = [#code]
def roll_dice(number):
# code
def get_terrain_by_time(terrain_id, time_id):
# code
def generate_terrain_description(terrain, time):
# code
def generate_terrain_image(description):
# code
def upload_base64_to_s3(bucket, key, base_64_str):
# code
def create_terrain(terrain_id, time_id, name, time, description, image_url):
dynamodb = boto3.resource('dynamodb')
table = dynamodb.Table('Terrains')
item = {
'terrain_id': terrain_id,
'time_id': time_id,
'terrain': name,
'time': time,
'description': description,
'image': image_url
}
response = table.put_item(Item=item)
if response['ResponseMetadata']['HTTPStatusCode'] == 200:
return item
else:
return None
def lambda_handler(event, context):
terrain_id = roll_dice(6)
time_id = roll_dice(2)
terrain = terrains_table[terrain_id -1]
time = time_table[time_id -1]
item = get_terrain_by_time(terrain_id, time_id)
if not item:
description = generate_terrain_description(terrain, time)
imageBase64 = generate_terrain_image(description)
item = create_terrain(terrain_id, time_id, terrain, time, description)
bucket = os.getenv("BUCKET")
key = f"terrains/{terrain}-{time}.png"
image_url = upload_base64_to_s3(bucket, key, imageBase64)
item = create_terrain(terrain_id, time_id, terrain, time, description, image_url)
return item
Here we just need to add the image_url property in the function definition and add a new image_url value in the item object.
Final result
If everything went well each time you test your Lambda function a new record will be stored in DynamoDB and store the image in S3, here is an example:
{
"terrain_id": 3,
"time_id": 2,
"description": "The night sky is illuminated by the stars and the moon, casting a pale light over the mountain range. The jagged peaks of the mountains are silhouetted against the night sky, creating a majestic and awe-inspiring sight. The air is crisp and cool, and the silence of the night is broken only by the occasional sound of a distant animal.",
"image": "https://{your-s3-bucket}.s3.amazonaws.com/terrains/Mountains-night.png",
"terrain": "Mountains",
"time": "night"
}
The image in this post was generated using the above criteria.
That's all for this post, in the next one probably I'll do some refactor as the code can get bigger, and it is better to have it in separate files.
Stay tuned!
Cover generated using OpenAI Dalle-2
Top comments (0)