Image generation

Storytell's Image Generation feature allows you to create visual content from textual descriptions. This feature directly translates natural language prompts into diverse visual outputs, enabling a wide range of creative and functional applications within the Storytell platform.

Written By Mark Ku

Last updated 3 months ago

⚠️ Important: Image generation is a beta feature and is more prone to errors like misspellings in charts and graphs. Make sure to double check the underlying data before using.

Key capabilities and underlying models

Storytell integrates with leading AI models to deliver image generation:

1. DALL-E 3 (OpenAI)

Best for: General-purpose, creative, and artistic images

Key Strengths:

  • Fast generation (produces images in seconds)

  • Versatile across many styles: photorealistic, illustrated, painted, abstract

  • Excels at vivid, dream-like, and surreal imagery

  • Great for marketing materials, advertisements, and social media

  • Handles complex, detailed prompts well

  • Ideal for business presentations and product visualization

Ideal use cases:

  • Creative and artistic content (illustrations, concept art)

  • Marketing and promotional materials

  • Quick iterations and exploring multiple creative directions

  • Product mockups and visualizations

2. Banana (Gemini 2.5 Flash Image)

Best for: Images with readable text and character consistency

Key Strengths:

  • #1 ranked on LMArena for image generation

  • High-fidelity text rendering (logos, signs, labels, diagrams)

  • Maintains character consistency across multiple images

  • Supports 10 different aspect ratios

  • Negative prompts to exclude unwanted elements

Ideal use cases:

  • Logos with text, book covers, and signage

  • Infographics and educational diagrams

  • Marketing materials combining imagery and typography

  • Brand assets requiring clear text

  • Character designs needing consistency across images

3. Imagen 3 (Google)

Best for: Photorealistic images

Key Strengths:

  • Superior photorealism with lifelike details

  • Exceptional texture rendering (materials, fabrics, surfaces)

  • Precise prompt adherence (follows instructions literally)

  • Professional-quality realistic imagery

Ideal use cases:

  • Product photography

  • Realistic portraits of people and animals

  • Architectural visualization and interior design

  • Nature and landscape photography

  • Food photography

  • Fashion and clothing photography

Implementation Details

When a user initiates an image generation request, Storytell translates the generic ImageGenerationRequest parameters into the specific API calls and configurations required by the chosen AI model (e.g., OpenAI's DALL-E 3 or Google's Imagen/Gemini). The system then processes the model's output, which might involve fetching image data from a temporary URL or decoding Base64 data. Any errors, such as a failed to generate image or an image being blocked by safety filters, are captured and communicated to ensure transparency and user awareness.