Dream Chaser Backend Server

A backend server for multilevel AI imaging.

Setup

1. Download pretrained text to image model from huggingface in subdirectory: git clone https://huggingface.co/prompthero/openjourney-v4
2. Download pretrained prompt generation model from huggingface in subdirectory: git clone https://huggingface.co/FredZhang7/distilgpt2-stable-diffusion-v2
3. Download tokenizer from huggingface in subdirectory: git clone https://huggingface.co/distilgpt2
4. Download pretrained sentence similarity model from huggingface in subdirectory: git clone https://huggingface.co/sentence-transformers/all-MiniLM-L6-v2

Run

1. Execute command: python start_api.py

API Requests

1. \image_gen: Generate an image using the AI text-to-image model. Returns a png object in bytes form.
   1. prompt (required): input text for image generation model
   2. height: height of image in pixels
   3. width: width of image in pixels
   4. num_inference_steps: number of denoising steps
   5. guidance_scale: guidance scale
   6. negative_prompt: prompts to avoid in image generation
   7. num_images_per_prompt: number of images to return
2. \image_upscale: Upscale an image using the AI upscaler model. Returns a png object in bytes form.
   1. url (required): location of input image for upscaler model
   2. prompt (required): input text for upscaler model
   3. num_inference_steps
   4. guidance_scale
   5. negative_prompt
   6. num_images_per_prompt
3. \text_gen: Improve a prompt using the AI text generation model. Returns a string.
   1. prompt (required): input text for text generation model
   2. temperature: diversity of prompt results
   3. top_k: number of tokens to sample at each step
   4. max_length: maximum number of output tokens
   5. repetition_penalty: penalty value for token repetition
   6. num_return_sequences: number of prompts to return