How to make AI Images of Yourself (Easy)

How to Create AI Images of Yourself in Five Steps

Step 1: Setting Up on Replicate

• The process begins with creating AI images quickly and affordably, emphasizing that anyone can do it.

• Users are instructed to visit replicate.com for training their face model and generating images. Signing in requires a GitHub account.

• A billing setup is necessary, costing approximately $2 for training the model and $0.05 per generated image.

Step 2: Training Your Face Model

• Navigate to the AI toolkit on replicate.com to start training your face model by selecting or creating a new model.

• Upload at least 12 high-quality images of your face, ensuring they are named with the trigger keyword "a photo of" followed by your name (e.g., "a photo of AI Andy").

• It’s crucial to use only the best photos; poor quality images may negatively impact the model's output.

Step 3: Creating a Hugging Face Account

• A Hugging Face account is required for obtaining an access token needed later in the process.

• After signing up, users must navigate to settings and create a new access token with appropriate permissions (excluding billing).

• Users should also create a new public model on Hugging Face, which will be linked back during training.

Step 4: Training Completion

• Once all values are set correctly, including linking Hugging Face repo ID, training will commence and typically takes about 30 minutes.

• Progress can be monitored via the dashboard where users can see logs from 0% to 100%, confirming successful completion once finished.

Step 5: Generating Images

• To generate an image using the trained model, users should go to a specific link provided in the description.

• The speaker discusses alternatives to Google search for finding AI tools more effectively without ads by using Brave Search instead.

How to Create AI Images of Your Face

Setting Up Image Generation Parameters

• The process begins with selecting the aspect ratio (16x9) and determining the number of outputs. A higher number of inference steps (recommended: 28) results in better quality images, while lower settings yield faster but blurrier results.

• Users can specify hidden inputs such as output format (JPEG) and quality settings (100). After configuring these options, clicking "run" initiates image generation.

Custom Prompting for Better Results

• To enhance image quality, it's suggested to create a custom prompt using ChatGPT instead of writing a lengthy mega prompt. This involves exploring GPT options and creating or editing an existing model.

• The user is encouraged to modify trigger keywords within the prompt configuration to personalize their experience. This customization aims to make prompts more visually appealing through effective prompt engineering principles.

Focusing on Personal Features

• When re-entering the same prompt, users should emphasize their facial features by adding descriptors like "closeup of face." This adjustment helps ensure that generated images focus on the user's likeness.

• Users can download generated images directly or as a zip file containing multiple outputs. The importance of experimenting with different prompts is highlighted for varied results.

Generating Multiple Outputs Simultaneously

• For efficiency, users can open new tabs to generate multiple images at once by copying prompts into separate links while ensuring all configurations are checked.

• Examples include creative prompts like "AI NDS as Iron Man," showcasing how diverse themes can be explored through AI-generated imagery.

Reviewing Generated Images