TalkCodyTalkCody

Image Generator

TalkCody's AI image generation feature that creates high-quality images from text descriptions

What is Image Generation?

Image Generation is TalkCody's AI image generation feature that can create high-quality images based on your text descriptions. Whether it's concept art, illustrations, product prototypes, or creative images, AI can help you generate them quickly.

Supported Image Generation Models

TalkCody supports the following image generation models (Image Output):

ModelProviderFeatures
DALL-E 3OpenAIHigh quality, rich details, creative image generation
Nano Banana (Gemini 2.5 Flash Image)Google / OpenRouter / AI GatewayGoogle's latest image generation model, fast and efficient
Nano Banana Pro (Gemini 3 Pro Image)Google / OpenRouter / AI Gateway / OllamaGoogle's professional image generation, highest quality
Doubao SeedreamVolcEngineByteDance Doubao series, optimized for Chinese scenarios
GLM ImageZhipu AIZhipu GLM series, excellent Chinese understanding
Qwen Image MaxAlibabaTongyi Wanxiang, professional image generation

You need to configure the corresponding provider's API Key in API Keys Configuration to use the corresponding image generation model.

Configuration

Configure API Keys

Different image generation models require configuring the corresponding provider's API Key:

ModelRequired API Key
DALL-E 3OpenAI
Nano Banana / Nano Banana ProGoogle AI / OpenRouter / Vercel AI Gateway
Doubao SeedreamVolcEngine
GLM ImageZhipu AI
Qwen Image MaxAlibaba Qwen

For detailed configuration steps, please refer to API Keys Management.

How to Use

Use via Agent

  1. Click the agent selector in the top-left corner of the chat interface
  2. Select Image Generator agent
  3. Describe the image content you want
  4. AI will automatically generate the image

Call as Sub-agent

In conversations with other agents, you can call Image Generator to generate images:

User: Help me create a cover image about machine learning

AI: I'll call the image generation agent to create a cover image for you.
[Calling Image Generator agent]

Call Tool Directly

In agents that support Tools, you can directly enable the imageGeneration tool:

  1. Click the Tools icon above the input box
  2. Check the imageGeneration tool
  3. Request image generation in the conversation

Image Specifications

Size Options

SizeAspect RatioUse Cases
1024x10241:1Square images, social media avatars
1792x102416:9Landscape images, PPT covers, widescreen display
1024x17929:16Portrait images, phone wallpapers, story images

When generating PPTs, we recommend using 1792x1024 size, which is the standard 16:9 aspect ratio.

Quality Options

  • Standard - Standard quality, faster generation, lower cost
  • HD / High - High definition quality, richer details, suitable for professional use

Generation Count

Supports generating 1-4 images simultaneously, allowing you to choose the most satisfactory result.

Prompt Techniques

Effective Prompt Structure

[Subject Description] + [Environment/Background] + [Style/Artistic Effect] + [Lighting/Atmosphere] + [Technical Specifications]

Examples

Simple Prompt:

A cat on grass

Optimized Prompt:

An orange tabby cat sitting on a sunny green lawn, surrounded by wildflowers, realistic style, soft afternoon light, HD details, 8K quality

ElementExamples
Artistic StyleRealistic, cartoon, watercolor, oil painting, sketch
Lighting & AtmosphereSoft light, backlight, warm tones, cool tones, golden hour
Color SchemeVibrant colors, soft tones, monochrome, high contrast
Quality DescriptionHD, 8K, rich details, professional photography

Where Generated Images are Saved

Generated images are automatically saved in the application data directory:

~/Library/Application Support/com.talkcody/images/

Each image has a unique identifier and can be viewed and downloaded in the chat history.

Frequently Asked Questions

What to do if image generation fails?

  1. Check if the API Key is valid and has sufficient balance
  2. Confirm the current model supports image generation
  3. Check if the network connection is normal
  4. Try simplifying the prompt and avoid sensitive content

How to get better generation results?

  • Use detailed, specific descriptions
  • Clearly specify artistic style and lighting conditions
  • Refer to the prompt techniques above to optimize descriptions
  • Use HD quality option for finer results