AI Image Creation Process, OpenAI API DallE2

AI Art Production Process, OpenAI API DallE2 20230713

This is the process of creating AI art using image-generating AI. The person in charge (Prompter/Arranger) repeatedly uses the OpenAI API (Dall-E 2) to complete the AI art (image file). The reviewer (Reviewer) ensures that the AI art does not (and will not) infringe the copyright of other companies.

Worker (#of Swimlane: 4)

Outside Prompter
- Consider as a ‘Prompter, not necessarily a user of the workflow platform.
In-house Prompter
- The Arranger is also responsible for the subsequent arrangement of the images.
Arranger
- Responsible for arranging images. Specifically, the operation of re-creating images using prompts, editing designated areas using prompts, and creating image variations is repeated.
Reviewer
- Responsible for checking images. Specifically, if they feel that another company’s copyright has been infringed, they report the infringement.

Business Flow (Complexity of Workflow: 26)

1. input Prompt
- The Prompter inputs the PROMPT for image generation.
2. Arrange
- Arranger repeats image generation (PROMPT required), image editing (PROMPT required) and variation creation (PROMPT not required).
3. Review
- The Reviewer checks for infringements of other companies’ copyrights.
4. Publish (discard)
- Arranger terminates the whole process of image generation.
(prompt log+)
(prompt log+)
AreaEdit
- The workflow platform instructs the OpenAI API DallE2 to complete the image in transparent areas.
Generate four imgs
- The workflow platform instructs the OpenAI API DallE2 to generate four images.
Generate two images
- The workflow platform instructs the OpenAI API DallE2 to generate two images.
Retry2
- The workflow platform instructs the OpenAI API DallE2 to generate two additional images.
Retry4
- The workflow platform instructs OpenAI API DallE2 to generate four additional images.
Retry8
- The workflow platform instructs OpenAI API DallE2 to generate eight additional images.
Variation
- The workflow platform instructs the OpenAI API DallE2 to generate similar images.
\Shared drive\a1- Company-wide
- The workflow platform stores files on Google Drive.
24 prefix [end1]
27 prefix [end2]
randSet arranger
- The workflow platform nominates Arranger.

Business Process Variables (#of Data Item: 15)

Outside Prompter (email) q_Outside_Prompter__email_ 0
PROMPT (to Generate or AreaEdit) q_Prompt 1
- Beautiful Architecture on a white background, close-up photo, happy,
Prompt Outputs (←”PROMPT”) q_Prompt_Outputs 2
AreaEdit Source q_AreaEdit_Source 3
(AreaEdit Mask) q_AreaEdit_Mask 4
AreaEdit Outputs (←”PROMPT”+”AreaEdit”) q_AreaEdit_Outputs 5
Variation Source q_Variation_Source 6
Variation Outputs (←”Variation”) q_Variation_Outputs 7
(Output Backups) q__Output_Backups_ 8
activity note q_activity_note 9
activity log q_activity_log 13
Reviewer q_Reviewer 10
Artwork q_Artwork 11
((arranger)) q_arranger 12
- ${processInstance.initQuserId}
((activity time)) q_activity_time 14

Field Name, num, Initial Value

Download

ai-art-creation-process-20230713-en.qar
- ai-art-creation-process-20230713-en.csv (for Importing Data Item Definitions)

This archive contains the BPMN icons that are not available in the Basic edition.

Notes

It is a tool for so-called ‘Prompt Engineering’.
- Image generation AI (“text-to-image” technology for generating images from text) is called many times via the API.
This Workflow app will utilise OpenAI’s Dall-E 2.
- OpenAI: Images, Generate
  - https://support.questetra.com/addons/openai-images-generate-2023/
- OpenAI: Images, Edit Specific Area
  - https://support.questetra.com/addons/openai-images-edit-specific-area-2023/
- OpenAI: Images, Create Variation
  - https://support.questetra.com/addons/openai-images-create-variation-2023/
- English prompt settings are recommended.
A 1024×1024 PNG image is generated.
- Each iterative operation produces between 2 and 8 images.
A separate contract for the use of the OpenAI API is required.
- A Secret API Key needs to be set prior to use.