Automate your Reels and TikToks Generation using Python

 Reference: https://c-nemri.medium.com/automate-your-reels-and-tiktoks-generation-using-python-faa4936159fe

In this blog, I guide you through automating your video content generation using python and generative AI. By the end of this blog, you’ll be capable of:

• Create TikTok videos and Instagram reels featuring your virtual avatar
• Automatically generating a video script by just providing a 1-line idea to your content pipeline using OpenAI API.
• Automatically generating voice over for your video script using ElevenLabs API
• Automatically generate videos where your virtual avatar says your video script using D-ID API

Register to ElevenLabs and Create your Voice

• Go to ElevenLabs website and register (Starter plan is recommended if you want to clone your voice)
• Follow the following steps to clone your voice

Register to D-ID and create your avatar

• Go to D-ID website and register (Lite plan is preferred to have a personal avatar)
• Follow the following steps to create your avatar

Register to OpenAI

• Register to OpenAI and create a new API key.

Automate your content pipeline using Python

Script generation

For script generation piece, we’ll use LangChain, which is a framework for developing applications powered by language models.

from langchain.llms import OpenAI
from langchain.prompts import PromptTemplate
from langchain.chains import LLMChain

llm = OpenAI(temperature=0)

template = """
You are the virtual avatar of Chouaieb Nemri. Given the following idea, write a 1 minute tiktok video script. Only write spoken words, not visuals. The speaker will be an AI generated avatar of myself.
Idea: {idea}
Script:
"""

prompt = PromptTemplate(
    template=template,
    input_variables=["idea"],
)

idea = """
Best Machine Learning course Andrew Ng's Machine Learning Specialization on Coursera
"""

chain = LLMChain(llm=llm, prompt=prompt)
script = chain.run(idea)
print(f"Generated script: {script}")
print("Is this script good? (y/n)")
user_input = input()
while user_input.lower() != "y":
    script = chain.run(idea)
    print(f"Generated script: {script}")
    print("Is this script good? (y/n)")
    user_input = input()

Audio Generation

We’ll use ElevenLabs python SDK to easily interact with their API. Code will look like the following.

from elevenlabs import generate, play, set_api_key, voices

voice_name = "Chouaieb"

print(f"Looking for voice id of {voice_name}")
voice_id = list(filter(lambda voice: voice.name == "Chouaieb", voices()))[0].voice_id

print(f"Found voice id: {voice_id}")
print(f"Generating audio with voice id: {voice_id}")

audio = generate(text=script, voice=voice_id, model="eleven_monolingual_v1")

with open("audio.mp3", "wb") as f:
    f.write(audio)

Uploading Audio and avatar to Amazon S3

We’ll have to upload audio and avatar image to Amazon S3 and share the objects using Pre-signed URLs so that D-ID is able to access them and generate the video.

import boto3

bucket_name = "<YOUR BUCKET NAME>"
audio_file = "<YOUR AUDIO FILE NAME>"
image_file = "<YOUR IMAGE FILE NAME>"

session = boto3.Session(profile_name="personal")
s3_client = session.client('s3')
s3_client.upload_file(image_file, bucket_name, image_file)
s3_client.upload_file(audio_file, bucket_name, audio_file)

audio_url = s3_client.generate_presigned_url(
    ClientMethod='get_object',
    Params={
        'Bucket': bucket_name,
        'Key': audio_file
    }
)

image_url = s3_client.generate_presigned_url(
    ClientMethod='get_object',
    Params={
        'Bucket': bucket_name,
        'Key': image_file
    }
)

Generating and Downloading Video

We’ll use D-ID API to generate the video. The following code launches a video creation job, waits for it until it finishes, then downloads the video.

import requests
import datetime

url = "https://api.d-id.com/talks"

payload = {
    "script": {
        "type": "audio",
        "reduce_noise": "false",
        "audio_url": audio_url
    },
    "source_url": image_url,
    "name": f"tiktok-video-{datetime.datetime.now().strftime('%Y-%m-%d-%H-%M-%S')}",
    "config": {
        "stitch": True
    }
}

headers = {
    "accept": "application/json",
    "content-type": "application/json",
    "authorization": "Basic <YOUR D-ID API KEY>"
}

print("Creating video...")

video_creation_response = requests.post(url, json=payload, headers=headers)
video_id = video_creation_response.json()['id']

url = f"https://api.d-id.com/talks/{video_id}"

video_get_response = requests.get(url, headers=headers)
while video_get_response.json()["status"] != "done":
    video_get_response = video_get_response.get(url, headers=headers)

video_url = video_get_response.json()["result_url"]

r = requests.get(result_url, allow_redirects=True)
open("video.mp4", 'wb').write(r.content)