AI enhanced prompts for 'dall-e-3', code cleanup

README.md

@@ -9,7 +9,7 @@ This is not affiliated with OpenAI in any way, and no OpenAI API key is required
 
 Compatibility:
 - [x] generations
-- - [x] prompt
+- - [x] prompt, including AI enhanced prompts when selecting 'dall-e-3'
 - - [x] model
 - - [x] size
 - - [ ] quality ("hd or standard")
@@ -20,19 +20,27 @@ Compatibility:
 - [ ] edits
 - [ ] variations
 
-This is in active development and currently works for basic use-cases. It selects generation configuration based on model.
+This is in active development and currently works for basic use-cases. It selects the generation configuration based on the chosen model.
 
-Currently hard coded as:
+The default configuration uses:
 - dall-e-1: sd 1.5
 - dall-e-2: sdxl_lightning
 - dall-e-3: sdxl
 
+Settings for each model type can be found in the config folder, and can be modified as needed without needing to restart the server.
+
 Quickstart
 ----------
 
-Docker (**recommended**):
+Configure your environment:
 ```shell
 echo SD_BASE_URL=http://<your stable diffusion webui server>:<port> > .env
+echo OPENAI_BASE_URL=http://<your openai chat server>:<port>/v1 >> .env # optional, for dall-e-3 style AI prompt enhancement
+echo OPENAI_API_KEY=skip >> .env # optional, for dall-e-3 style AI prompt enhancement
+```
+
+Docker (**recommended**):
+```shell
 docker compose up
 ```
 or:
@@ -41,14 +49,18 @@ pip install -r requirements.txt
 python images.py --host 127.0.0.1 --port 5005
 ```
 
-You can use the OpenAI client to interact with the API.
+You can use the OpenAI python client to interact with the API.
+```shell
+pip install -U openai
+```
+
 ```python
 import base64
 import io
 import openai
 from PIL import Image
-client = openai.Client(base_url='http://localhost:5005/v1')
-response = client.images.generate(prompt="A cute baby sea otter", response_format='b64_json')
+client = openai.Client(base_url='http://localhost:5005/v1', api_key='skip')
+response = client.images.generate(prompt="A cute baby sea otter", response_format='b64_json', model='dall-e-3')
 image = Image.open(io.BytesIO(base64.b64decode(response.data[0].b64_json)))
 image.show()
 
@@ -61,4 +73,5 @@ Links & Documentation
 - OpenAI Images Guide: (https://platform.openai.com/docs/guides/moderation)
 - OpenAI Images API Reference: (https://platform.openai.com/docs/api-reference/moderations)
 - AUTOMATIC1111 Stable Diffusion WebUI: (https://github.com/AUTOMATIC1111/stable-diffusion-webui)
-- ashleykleynhans/stable-diffusion-docker (https://github.com/ashleykleynhans/stable-diffusion-docker)
+- ashleykleynhans/stable-diffusion-docker (https://github.com/ashleykleynhans/stable-diffusion-docker)
+

config/default_sd15_conf.json

@@ -1,4 +1,12 @@
 {
   "sampler_name": "DPM++ 2M Karras",
-  "steps": 30
+  "steps": 30,
+  "hr_upscaler": "Latent",
+  "denoising_strength": 0.68,
+  "subseed_strength": 0.75,
+  "override_settings": {
+    "sd_model_checkpoint": "realisticVisionV60B1_v51VAE.safetensors",
+    "sd_vae": "vae-ft-mse-840000-ema-pruned.ckpt"
+  },
+  "override_settings_restore_afterwards": false
 }

config/default_sdxl_conf.json

@@ -2,6 +2,13 @@
   "sampler_name": "DPM++ 2S a Karras",
   "refiner_checkpoint": "sd_xl_refiner_1.0.safetensors",
   "refiner_switch_at": 0.8,
-  "VAE": "fixFP16ErrorsSDXLLowerMemoryUse_v10.safetensors",
-  "steps": 40
+  "steps": 40,
+  "hr_upscaler": "Latent",
+  "denoising_strength": 0.68,
+  "subseed_strength": 0.75,
+  "override_settings": {
+    "sd_model_checkpoint": "opendalle_v11.safetensors",
+    "sd_vae": "fixFP16ErrorsSDXLLowerMemoryUse_v10.safetensors"
+  },
+  "override_settings_restore_afterwards": false
 }

config/default_sdxl_lightning_conf.json

@@ -2,7 +2,14 @@
   "sampler_name": "DPM++ SDE",
   "refiner_checkpoint": null,
   "refiner_switch_at": 0.0,
-  "VAE": null,
   "steps": 4,
-  "cfg_scale": 1.0
-}
+  "cfg_scale": 1.0,
+  "hr_upscaler": "Latent",
+  "denoising_strength": 0.68,
+  "subseed_strength": 0.75,
+  "override_settings": {
+    "sd_model_checkpoint": "dreamshaperXL_lightningDPMSDE.safetensors",
+    "sd_vae": null
+  },
+  "override_settings_restore_afterwards": false
+}

docker-compose.yml

@@ -2,9 +2,12 @@ services:
   server:
     build:
       dockerfile: Dockerfile
+    tty: true
     image: ghcr.io/matatonic/openedai-images
     env_file:
       - .env
+    volumes:
+      - ./config:/app/config
     ports:
       - 5005:5005
     command: ["python", "images.py", "--host", "0.0.0.0", "--port", "5005"]