Video 1 (Link)

Notes:

Overview

The video presents a guide on generating photorealistic human portraits using the free AI image generator "Tensor Art," highlighting specific models, key settings, and a detailed prompt for optimal results.

Introduction to Tensor Art

• Tensor Art is a free tool that generates various image styles, including cartoons, anime, photorealistic, and fantasy art.
• The model recommended for realistic portraits is called "Juggernaut Aftermath."

Using the Juggernaut Aftermath Model

• Users must select the base model (defaulting to Juggernaut Aftermath) but may choose others if desired.
• Optional settings allow adding LoRA (Low-Rank Adaptation) models or control nets, though default settings are advised for realism.
• There are options to input positive prompts, negative prompts, and to upload a reference image.
• Aspect ratios (portrait, landscape, square, or custom) can be selected.
• Advanced settings like sampling methods and seed values are available but not required for typical use.

Prompt Strategies for Photorealism

• Initial prompt example: "Beautiful woman with curly hairs close up portrait realistic," with negatives like "no blur, ugly face, extra hands, extra eyes."
• Results were realistic but not optimal; better results require more detailed prompts.
• Recommended "magical" prompt: "A beautiful lady, freckles, big smile, dark makeup, hyper detailed photography, head and shoulders portrait, cover photo."
• Critical negative prompt: "Worst quality, low quality, normal quality, low details, oversaturated, undersaturated, overexposed, underexposed."
• Combining these prompts yields highly realistic images with detailed textures.

Experimenting with Variations

• Changing the positive prompt to include contexts like "sitting in a coffee shop" produces convincing backgrounds and maintains photorealism.
• Adjusting seed values allows further image variety with consistently realistic results.

Advantages and Recommendations

• Tensor Art simplifies the process compared to installing Stable Diffusion and LoRA models locally, requiring no advanced hardware.
• The provided prompts significantly enhance output quality from free AI image generators.

Action Items

• TBD – Viewers: Use the detailed positive and negative prompts to generate photorealistic portraits on Tensor Art.
• TBD – Content Creator: Share the exact prompt texts in the video description for viewer reference.

Questions / Follow-Ups

• How do other free AI image generators compare when using similar detailed prompts?
• Are there additional models on Tensor Art that surpass Juggernaut Aftermath for specific use cases?

Video 2 (Link)

Notes:

Overview

This guide outlines a five-step framework for generating realistic images with ChatGPT’s image capabilities by crafting effective prompts.

Step 1: Define Subject and Setting

• Begin with a clear, specific subject and setting.
• Provide detailed context to reduce ambiguity for the model.

Step 2: Add Descriptive Details

• Include details such as facial expressions, textures, clothing, age, and mood.
• Aim to create a vivid and precise mental image.

Step 3: Specify Technical Camera Details

• Mention camera models, lenses, and focal lengths to influence depth of field and perspective.
• For shallow depth of field, specify something like an 85mm lens at f1.2.
• For a wide cinematic effect, suggest a 24mm lens.
• Providing camera and lens details offers visual context.

Step 4: Describe Lighting

• Clearly state lighting conditions, such as soft daylight for a natural look or harsh backlight for drama.
• Note light direction, softness, and source to shape the atmosphere.

Step 5: Guide Composition and Style

• Indicate subject framing (centered, off-center, rule of thirds).
• Reference artistic styles, camera movements, or moods as needed to refine the image.

Recommendations / Advice

• Combine all five elements in your prompt for the most realistic results.
• Experiment with different details to see how each aspect affects the generated image.

Video 3 (Link)

Notes:

Overview

The video reviews five leading AI image generators, explaining their features, free usage limits, and performance under similar testing conditions. A side-by-side comparison highlights strengths, weaknesses, and ideal use cases for each tool.

Rev (preview.rev.art)

• Free account gives 100 initial credits and 20 daily credits, each image costs 1 credit.
• Simple interface; no predefined styles, users must specify desired style in prompt.
• Can enhance prompts for more detailed results; aspect ratio adjustable.
• Supports generating up to eight images at once; “Instruct” lets users alter parts of an image.
• Image-to-image feature copies aspect ratio but doesn’t use uploaded prompts.
• Best at realistic close-up portraits, but struggles with faces at distance.

Runway

• Known for video creation, but also provides image generation (≈15 free images).
• Offers predefined styles, aspect ratio adjustment, aesthetic range, and reference images.
• Reference images enable consistent character generation; can save and reuse named characters.
• Can generate variations and turn images into videos.
• Produces very consistent characters from references; good with complex prompts and real objects/brands.
• Blocked some generations due to policy restrictions.

Google Image Generators (ImageFX & Wisk)

• ImageFX requires VPN in some regions; allows style/effect keywords and seed locking for character consistency.
• Detailed prompts and locked seeds help achieve similar characters.
• Wisk adds support for object, environment, and style references; Imagin 4 model available.
• Strong censorship, especially with images of people.
• Realism and consistency vary; reference images less effective for exact character replication.
• Struggled with complex hand prompts; otherwise offers generous free plans.

Ideogram AI

• Provides 10 weekly credits; latest model offers higher quality at higher credit cost.

Video 4 (Link)

Video 5 (Link)