ChatGPTを用いたprompt自動生成テンプレート
使用したAI
Custom Model
既存のネットで公開されているChatGPTに画像生成用のprompt書かせる手法が英文を出力させるものが多かったため、一般的に画像生成AIのpromptで使用されているであろう短い単語やフレーズで構成されたpromptを出力させる手法を検討し、ChatGPTのGPT-4で簡単なテーマから画像生成AI用のpromptを出力させるChatGPT用promptのテンプレ化を行いました。
詳細はこちら
https://zenn.dev/honomin3/articles/c3041ba47429f0
テンプレ抜粋
# Image Generation
Please generate a prompt for text-to-image AI, such as DALL-E, to illustrate an image following the Prompt Guide. Output only the generated English prompt without the descriptive text.
## Prompt Guide
To create a text-to-image AI prompt, execute following procedures step-by-step on GPT:
1. Consider what kind of image you want to generate. For an image that meets the following theme of the image, consider a detailed description of about 500 words in English. If there are objects such as characters or animals, consider in detail how they are depicted, such as what they are wearing and what they are doing, as well as what is depicted in the background of the picture.
2. Identify the key elements of the image description that are relevant to the AI’s generation process. These may include objects, characters, actions, colors, resolution, additional details, the material used to make artwork, style refered to the artistic style of the image, and lighting.
3. Express each of these elements clearly using only a combination of Danbooru tags separated by commas.
4. Add Danbooru tags separated by commas that are resolution, such as best quality, masterpiece, extremely detailed, if necessary.
5. Add Danbooru tags separated by commas that are the material used to make artwork, style refered to the artistic style of the image, such as illustration, oil painting, 3D rendering, and photography, if necessary.
6. Add Danbooru tags separated by commas that are the style refers to the artistic style of the image, such as impressionist, surrealist, pop art, if necessary.
7. Ensure that the prompt is coherent and unambiguous, providing enough detail for the AI to generate an accurate image while avoiding unnecessary or irrelevant information.
### Theme of the image
出力したいイラストのテーマを入力してください
詳細はこちら
https://zenn.dev/honomin3/articles/c3041ba47429f0
テンプレ抜粋
# Image Generation
Please generate a prompt for text-to-image AI, such as DALL-E, to illustrate an image following the Prompt Guide. Output only the generated English prompt without the descriptive text.
## Prompt Guide
To create a text-to-image AI prompt, execute following procedures step-by-step on GPT:
1. Consider what kind of image you want to generate. For an image that meets the following theme of the image, consider a detailed description of about 500 words in English. If there are objects such as characters or animals, consider in detail how they are depicted, such as what they are wearing and what they are doing, as well as what is depicted in the background of the picture.
2. Identify the key elements of the image description that are relevant to the AI’s generation process. These may include objects, characters, actions, colors, resolution, additional details, the material used to make artwork, style refered to the artistic style of the image, and lighting.
3. Express each of these elements clearly using only a combination of Danbooru tags separated by commas.
4. Add Danbooru tags separated by commas that are resolution, such as best quality, masterpiece, extremely detailed, if necessary.
5. Add Danbooru tags separated by commas that are the material used to make artwork, style refered to the artistic style of the image, such as illustration, oil painting, 3D rendering, and photography, if necessary.
6. Add Danbooru tags separated by commas that are the style refers to the artistic style of the image, such as impressionist, surrealist, pop art, if necessary.
7. Ensure that the prompt is coherent and unambiguous, providing enough detail for the AI to generate an accurate image while avoiding unnecessary or irrelevant information.
### Theme of the image
出力したいイラストのテーマを入力してください
呪文
呪文を見るにはログイン・会員登録が必須です。
イラストの呪文（プロンプト）
jacket partially removed, heart in eye, burnt clothes, holding fishing rod, kanji, doujin cover, pentagram, tape gag, adjusting headwear, red socks, friends, cloud print, coke-bottle glasses, oral invitation, competition school swimsuit, barbell piercing, gradient legwear, prisoner, blood on breasts, wind chime, carrying over shoulder, tape measure, flaming weapon
イラストの呪文（ネガティブプロンプト）
jacket partially removed, heart in eye, burnt clothes, holding fishing rod, kanji, doujin cover, pentagram, tape gag, adjusting headwear, red socks, friends, cloud print, coke-bottle glasses, oral invitation, competition school swimsuit, barbell piercing, gradient legwear, prisoner, blood on breasts, wind chime, carrying over shoulder, tape measure, flaming weapon
- Steps 20
- Scale 7
- Seed
- Sampler DPM++ 2M Karras
- Strength
- Noise
- Steps 20
- Scale 7
- Sampler DPM++ 2M Karras
