ai-studio-image

## Overview

Humanized image generation via Google AI Studio (Gemini). Realistic influencer-style or educational photos with natural lighting and subtle imperfections.

## When to Use This Skill

- When the user mentions "generate an image" or related topics
- When the user mentions "generate a photo" or related topics
- When the user mentions "create an image" or related topics
- When the user mentions "realistic photo" or related topics
- When the user mentions "humanized image" or related topics
- When the user mentions "influencer photo" or related topics

## Do Not Use This Skill When

- The task is unrelated to ai studio image
- A simpler, more specific tool can handle the request
- The user needs general-purpose assistance without domain expertise

## How It Works

The difference between an AI image and a real photo lies in the imperceptible details:
the slight grain of a phone sensor, the imperfect lighting, the framing
that is slightly off-center, the depth of field characteristic of a small lens.
This skill systematically injects those qualities into every generation.

## Ai Studio Image — Humanized Image Specialist

An image-generation skill via Google AI Studio that turns any prompt into photos
with a genuinely human look. Every generated image looks like it was taken by a
real person with their phone — not by an AI.

## 1. Configure The Api Key

The user needs a Google AI Studio API key:
- Go to https://aistudio.google.com/apikey
- Create or copy your API key
- Set it as an environment variable:

```bash

## Windows

set GEMINI_API_KEY=your-api-key-here

## Linux/Mac

export GEMINI_API_KEY=your-api-key-here
```

Or create a `.env` file in `C:\Users\renat\skills\ai-studio-image\`:
```
GEMINI_API_KEY=your-api-key-here
```

## 2. Install Dependencies

```bash
pip install -r C:\Users\renat\skills\ai-studio-image\scripts\requirements.txt
```

## 3. Generate Your First Image

```bash
python C:\Users\renat\skills\ai-studio-image\scripts\generate.py --prompt "young woman having coffee in a cafe" --mode influencer --format square
```

## Main Workflow

When the user asks to generate an image, follow this flow:

## Step 1: Identify The Mode

Ask or infer from context:

| Mode | When to Use | Characteristics |
|------|-------------|-----------------|
| **influencer** | Social media posts, lifestyle, personal branding | Attractive yet natural aesthetic, vibrant colors without oversaturation, attention-grabbing composition |
| **educational** | Course material, tutorial, presentation, infographic | Clean, professional look, focus on the content, clear and legible elements |

If the user doesn't specify, use **influencer** as the default for social-media content
and **educational** for anything related to teaching/presentation.

## Step 2: Identify The Format

| Format | Aspect Ratio | Ideal Use |
|---------|-------------|-----------|
| `square` | 1:1 | Instagram feed, Facebook, profiles |
| `portrait` | 3:4 | Instagram portrait, Pinterest |
| `landscape` | 16:9 | YouTube thumbnails, banners, desktop |
| `stories` | 9:16 | Instagram/Facebook Stories, TikTok, Reels |

If not specified, infer from context (stories → 9:16, feed → 1:1, etc).

## Step 3: Transform The Prompt

**This is the most important step.** Never send the user's prompt directly to the API.
Always pass it through the humanization engine:

```bash
python C:\Users\renat\skills\ai-studio-image\scripts\prompt_engine.py --prompt "user's prompt" --mode influencer
```

The humanization engine adds layers of realism:

**Layer 1 — Device and Technique:**
- Shot with a smartphone (iPhone/Samsung Galaxy)
- Phone lens with natural depth of field
- No flash — ambient light only
- Slight sensor noise (high ISO in low light)

**Layer 2 — Natural Lighting:**
- Indirect sunlight / golden hour / window light
- Soft, organic shadows
- No studio lighting
- Natural reflections on surfaces

**Layer 3 — Human Imperfections:**
- Slightly imperfect framing (not mathematically centered)
- Natural selective focus (something slightly out of focus in the background)
- Micro hand-tremor (sharpness isn't absolute)
- Random elements from the real environment

**Layer 4 — Authenticity:**
- Genuine facial expressions (not studio poses)
- Everyday clothes and settings
- Real skin texture (pores, subtle marks — no porcelain skin)
- Realistic body proportions

**Layer 5 — Environmental Context:**
- Real settings (not generic stock backgrounds)
- Everyday objects in the environment
- Lighting consistent with the setting
- Time of day coherent with the activity

## Step 4: Generate The Image

```bash
python C:\Users\renat\skills\ai-studio-image\scripts\generate.py \
  --prompt "humanized prompt generated in the previous step" \
  --mode influencer \
  --format square \
  --model gemini-2-flash-exp \
  --output C:\Users\renat\skills\ai-studio-image\data\outputs\
```

**Available models (in order of recommendation):**

| Model | Speed | Quality | Cost | Ideal Use |
|--------|-----------|-----------|-------|-----------|
| `gemini-2-flash-exp` | Fast | High | **FREE** | **Default — always use** |
| `imagen-4` | Medium | High | $0.03/img | High quality (requires --force-paid) |
| `imagen-4-ultra` | Slow | Maximum | $0.06/img | Print, 2K (requires --force-paid) |
| `imagen-4-fast` | Fast | Good | $0.02/img | High volume (requires --force-paid) |
| `gemini-flash-image` | Fast | High | $0.039/img | Image editing (requires --force-paid) |
| `gemini-pro-image` | Medium | Maximum+4K | $0.134/img | Reference, 4K (requires --force-paid) |

## Step 5: Present And Iterate

Show the result to the user. If adjustments are needed:
- Relight: Adjust the lighting
- Reframe: Change the composition
- More/less natural: Adjust the level of imperfections
- Change the setting: Alter the environment

## Pre-Configured Templates

For common scenarios, use ready-made templates. Run:

```bash
python C:\Users\renat\skills\ai-studio-image\scripts\templates.py --list
```

Available templates:

## Influencer Mode

| Template | Description |
|----------|-------------|
| `cafe-lifestyle` | Person in a cafe/restaurant with a drink/food |
| `outdoor-adventure` | Outdoor activity, nature, travel |
| `workspace-minimal` | Elegant work desk, home office |
| `fitness-natural` | Exercise/wellness with a natural look |
| `food-flat-lay` | Food seen from above, casual flat lay |
| `urban-street` | Urban setting, street style |
| `golden-hour-portrait` | Portrait with golden sunset light |
| `mirror-selfie` | Mirror selfie, casual and spontaneous |
| `product-in-use` | Product being used naturally by a person |
| `behind-scenes` | Behind the scenes, making-of, everyday life |

## Educational Mode

| Template | Description |
|----------|-------------|
| `tutorial-step` | Person demonstrating a tutorial step |
| `whiteboard-explain` | Person explaining at a board/whiteboard |
| `hands-on-demo` | Hands giving a practical demonstration |
| `before-after` | Before/after comparison |
| `tool-showcase` | Tool/software being used |
| `classroom-natural` | Classroom/workshop environment |
| `infographic-human` | Person pointing at data/charts |
| `interview-setup` | Natural interview/podcast setup |
| `screen-recording-human` | Person with a laptop showing the screen |
| `team-collaboration` | Team working together naturally |

To use a template:
```bash
python C:\Users\renat\skills\ai-studio-image\scripts\generate.py \
  --template cafe-lifestyle \
  --custom "redhead woman, 30 years old, reading a book" \
  --format square
```

## Humanization Level

Control how much "imperfection" to inject:

| Level | Effect |
|-------|--------|
| `ultra` | Maximum realism — looks 100% like a phone photo |
| `natural` (default) | Perfect balance between quality and realism |
| `polished` | Cleaner, still natural but with more aesthetic care |
| `editorial` | Magazine style, natural but with production value |

```bash
python C:\Users\renat\skills\ai-studio-image\scripts\generate.py \
  --prompt "..." --humanization natural
```

## Time Of Day

Lighting changes drastically:

| Option | Description |
|-------|-----------|
| `morning` | Soft morning light, cool-warm tones |
| `golden-hour` | Sunset/sunrise, golden tones |
| `midday` | Hard midday light, harsh shadows |
| `overcast` | Cloudy day, uniform diffuse light |
| `night` | Artificial lighting, warm tones |
| `indoor` | Interior light, mixed |

## Batch Generation

To generate multiple variations:

```bash
python C:\Users\renat\skills\ai-studio-image\scripts\generate.py \
  --prompt "..." --variations 4 --format square
```

## Instagram Skill

Generate images and publish directly:
1. Use `ai-studio-image` to generate the photo
2. Use the `instagram` skill to publish with an optimized caption

## Canva Integration

Generated images can be sent to Canva to add text/branding.

## Troubleshooting

| Problem | Solution |
|----------|---------|
| `GEMINI_API_KEY not found` | Set the environment variable or create `.env` |
| `quota exceeded` | Wait for the rate limit to reset or upgrade the plan |
| `image blocked` | Adjust the prompt — it may contain restricted content |
| `low quality output` | Raise humanization to `ultra`, try another model |

## References

For detailed guides, see:
- `references/setup-guide.md` — Full installation and configuration
- `references/prompt-engineering.md` — Advanced prompt techniques for humanized images
- `references/api-reference.md` — Google AI Studio API documentation

## Best Practices

- Provide clear, specific context about your project and requirements
- Review all suggestions before applying them to production code
- Combine with other complementary skills for comprehensive analysis

## Common Pitfalls

- Using this skill for tasks outside its domain expertise
- Applying recommendations without understanding your specific context
- Not providing enough project context for accurate analysis

## Related Skills

- `comfyui-gateway` - Complementary skill for enhanced analysis
- `image-studio` - Complementary skill for enhanced analysis
- `stability-ai` - Complementary skill for enhanced analysis