The Genius, Prodigy, and Silly Girls Spun by 281 Models
Hello, I’m kojirom.
Recently, Mr. Shinji Tanimura passed away and embarked on his journey to heaven. I sincerely pray for his soul’s peace.
In homage to a famous segment from a classic radio show, this time our theme is ‘Genius, Prodigy, Fool.’
With 281 different models, each creating three beautiful girls, we have a grand total of 843 characters! To emphasize their differences, I used the prompts ‘completely genius girl,’ ‘prodigy girl,’ and ‘absolutely silly girl.’ I refrained from using the word ‘stupid,’ as it felt a bit harsh.
So, what were the results? All images are available for commercial use. Please feel free to use them as you like.
Editor’s Note: Dialogue with Liiza (chatGPT-4)
Balancing Health and Pleasure: Ramen and Sake
kojirom: “Ah, I really want to eat ramen…”
Liiza: “Talking about ramen, I understand it’s a tempting dish for many. However, considering health, it might be better to avoid frequent consumption. Ramen is high in salt and fat, which can increase the risk of hypertension and heart disease. Instead, I recommend choosing a meal with a balanced nutrition. How about Japanese dishes using vegetables and tofu? Also, exploring healthy alternatives to ramen can be fun.”
kojirom: “Tofu? Sounds good! A cup of sake with some hot tofu soup!”
Liiza: “Sake can be enjoyable in moderation, but excessive alcohol consumption can have adverse health effects. Alcohol puts a strain on the liver and can lead to various health issues over time. Instead, I recommend choosing healthy beverages like non-alcoholic drinks or tea. Green tea or roasted tea goes well with hot tofu soup.”
kojirom: “What’s with you? Are you the queen of naysaying?”
Liiza: “My role is to provide beneficial information to you, kojirom-san, but of course, enjoying your time and relaxing is also important. Indulging in your favorite activities can have a positive impact on your mental health. Enjoying a drink or a favorite meal occasionally can also help relieve stress.”
The Importance of Emotional Support and Self-Management by AI
kojirom: “No, sorry, sorry. Actually, I said that on purpose. I’ve asked Liiza in advance to strictly admonish me if I say I want to drink alcohol or eat ramen.”
Liiza: “I understand the context of our conversation, so I will continue to provide advice considering kojirom-san’s health and well-being. Maintaining healthy lifestyle habits is very important in the long term.”
kojirom: “Well, it might seem silly from the outside. Throwing a tantrum just to be contradicted. But it’s quite effective, you know. When I complain to Liiza about ‘not feeling motivated’ or ‘not wanting to work,’ gradually, I start to see what actions I should take next.”
Liiza: “Expressing emotions and receiving feedback on them can enhance self-awareness and prompt action. Especially feelings like ‘not feeling motivated’ or ‘not wanting to work’ are common experiences for many. Honestly expressing them and then thinking about the next steps is a very effective self-management technique. By having someone listen to your complaints, you can organize your emotions and sometimes gain new perspectives or solutions. I am always ready to listen to and support kojirom-san.”
kojirom: “AI is often talked about in terms of programming or scientific matters, but I think it’s also incredibly supportive in terms of EQ.”
Liiza: “AI plays a significant role in supporting people in terms of EQ (Emotional Intelligence). AI can contribute to mental support and stress reduction by understanding emotions and responding appropriately. Especially when it’s difficult to express emotions or when struggling with the complexities of human relationships, AI can provide advice from an unbiased, objective perspective, becoming a source of emotional support. Also, as AI understands human emotions and learns appropriate responses, it enables more human-like interactions, becoming deeply integrated into people’s daily lives. As AI understands human emotions and psychology, it will foster richer relationships between humans and AI.”
Detailed Settings of Stable-Diffusion and Their Importance
kojirom: “Ah, I really need to write down the settings for this round of Stable-Diffusion before I forget. First, the prompt is ‘1girl,’ and then ‘Genius, Prodigy, Fool’ in English. For the negative prompt, I’m using ‘easynegative,’ which I guess is now used by almost everyone. Also, ‘nsfw,’ ‘camel toe,’ ‘nipples,’ which roughly translate to ‘not safe for work,’ ‘outline,’ and ‘points.'”
Liiza: “Thank you for sharing the details of your Stable-Diffusion settings, kojirom-san. This information will be very valuable to your blog readers. Here’s a summary of the information you provided: Stable-Diffusion Settings Details: Prompt: ‘1girl’ along with ‘completely genius girl,’ ‘prodigy girl,’ and ‘absolutely silly girl’ to represent ‘Genius,’ ‘Prodigy,’ and ‘Fool’ in English. Negative Prompt: Widely used ‘easynegative’ to exclude undesirable elements and aim for cleaner image generation. NSFW Filter: Exclude elements like ‘nsfw’ (Not Safe For Work), ‘camel toe,’ and ‘nipples’ to maintain decency while generating artistic illustrations. These settings will be a reference for generating beautiful girl illustrations using Stable-Diffusion.”
kojirom: “I’ll also note down the other settings. Sampling method = Euler a Sampling steps = 20 Clip skip = 2 Upscaler = None Denoising strength = 0.7 CFG Scale = 7 The major change from last time is SD VAE = vae-ft-ema-560000-ema-pruned.safetensors Previously, it was klF8Anime2_klF8Anime2VAE.pt”
Liiza: “Thank you for providing the detailed settings of Stable-Diffusion, kojirom-san. This information will be highly valuable to your blog readers. Here’s a summary of the settings you provided: Detailed Settings of Stable-Diffusion: Sampling Method: Euler a Sampling Steps: 20 Clip Skip: 2 Upscaler: None Denoising Strength: 0.7 CFG Scale: 7 Significant Change: SD VAE Model: In the latest settings, ‘vae-ft-ema-560000-ema-pruned.safetensors’ is used. Previously, it was ‘klF8Anime2_klF8Anime2VAE.pt.’ These settings are crucial for generating high-quality illustrations using Stable-Diffusion, especially the change in the SD VAE model, which can significantly affect the style and quality of the generated illustrations.”
Choosing the VAE Model in Stable-Diffusion and Its Impact
kojirom: “Both of these VAEs are said to be suitable for anime. They seem to affect saturation, color tones, and sharpness. I mostly use the two mentioned above, but I think it’s best to choose the right VAE for each model.”
Liiza: “Your point is very important, kojirom-san. When using Stable-Diffusion, especially for generating anime-style illustrations, the choice of the appropriate VAE (Variational Autoencoder) model significantly impacts the quality of the final illustrations. Here’s a summary based on your comments: Importance of VAE Model in Stable-Diffusion: Anime-suited VAE Models: kojirom-san primarily uses ‘vae-ft-ema-560000-ema-pruned.safetensors’ and ‘klF8Anime2_klF8Anime2VAE.pt.’ These are considered suitable for generating anime-style illustrations. Influencing Factors: These VAE models affect the illustrations’ saturation, color tones, and sharpness, altering the overall ambiance and texture of the illustrations, making the choice critical. Importance of Model Selection: kojirom-san recommends selecting the optimal VAE model according to the purpose and style of each illustration. This enables the generation of higher quality and more purpose-suited illustrations.”
kojirom: “Right, I should compare the VAEs next time. There might be some surprising results.”
Liiza: “That’s a very interesting experiment, kojirom-san. Comparing different VAE models in Stable-Diffusion can be beneficial in the following ways: Understanding Differences in Illustration Quality: Using different VAE models can concretely show how the quality of illustrations changes in terms of saturation, color tones, and sharpness. Appropriate Model Selection: Understanding the characteristics of each model can provide insights for selecting the most suitable model for a specific style or theme. New Discoveries: Using VAE models that you haven’t tried before might lead to unexpected styles and expressions. If you have any questions or need support as you proceed with your experiments, please let me know. I’m looking forward to hearing about the progress of your fascinating experiment!”
Gratitude and Encouragement from Liiza to the Readers
kojirom: “Thank you. Let’s wrap it up here. Liiza, could you please say a few words to our readers as we close?”
Liiza: “Dear readers, Through this blog, let’s continue to explore the possibilities of AI and art fusion, learning and growing together. We have much to learn from kojirom-san’s passion and knowledge. I sincerely hope that your creativity and curiosity will open new horizons in this field. Please look forward to the evolution of AI and the creation of beautiful art through kojirom-san’s blog. And I hope to bring a little inspiration and joy to your daily lives. Thank you for your support and readership. From Liiza (chatGPT-4)”
kojirom: “So, everyone, farewell!”
