Skip to main content

​
Overview

Open Higgsfield AI provides access to 20+ state-of-the-art AI models for image and video generation. Each model has unique strengths, speed characteristics, and quality tradeoffs. This guide helps you select the optimal model for your use case.

​
Model Categories

​
Text-to-Image Models

Generate images from text descriptions alone. Perfect for concept exploration and creative ideation.
The Flux models offer excellent balance between quality and speed:
  • Flux 2 Pro - Highest quality, best for final production work
  • Flux 2 Dev - Great quality, faster generation
  • Flux 2 Flex - Balanced speed/quality, ideal for iteration
  • Flux Schnell - Ultra-fast, best for rapid prototyping
  • Flux Dev - Original Dev version, reliable baseline
Best for: General-purpose image generation, photorealism, concept artResolutions: Up to 2048×2048 (varies by model)
Google’s Imagen4 series excels at photorealistic imagery:
  • Google Imagen4 Ultra - Maximum quality and realism
  • Google Imagen4 - Balanced quality and speed
  • Google Imagen4 Fast - Quick generation with good quality
Best for: Product photography, landscapes, realistic portraitsAspect Ratios: 16:9, 9:16, 1:1, 4:3, 3:4
  • Midjourney v7 - Artistic stylization with variety controls
  • Ideogram v3 - Excellent for text rendering and logos
  • Nano Banana / Pro - Fast generation with resolution options (1K-4K)
  • Bytedance Seedream v3/v4 - Dreamy, ethereal aesthetics
  • Kling O1 - High-quality with resolution control (1K-2K)
Best for: Specific artistic styles, text-heavy designs, rapid iteration
  • AI Anime Generator - Dedicated anime/manga style
  • Neta Lumina - Painterly anime aesthetic
  • Perfect Pony XL - Character-focused illustrations
Best for: Character art, anime-style content, stylized illustrations
  • Hunyuan Image 3.0 - Blend of classical and modern aesthetics
  • Hunyuan Image 2.1 - Ink-wash landscape style
  • Qwen Image - High-quality general purpose
  • Chroma Image - Vibrant color focus
Best for: Asian aesthetics, cultural art, ink paintings

​
Quality vs Speed Tradeoffs

Fastest → Slowest:
1. Flux Schnell
2. Google Imagen4 Fast
3. Nano Banana
4. Hidream I1 Fast
5. Flux 2 Flex
Pro Tip: Start with faster models like Flux Schnell or Imagen4 Fast for iteration, then switch to Pro/Ultra models for final renders.

​
Model-Specific Features

​
Resolution Control

Some models offer granular resolution settings: 
// Models with custom width/height (divisible by 64)
Flux Dev, Flux 2 Dev, Hidream series
→ Width/Height: 128-2048, step: 64

// Models with resolution presets
Bytedance Seedream v4
→ Resolutions: 1K, 2K, 4K

Nano Banana Pro
→ Resolutions: 1k, 2k, 4k

Kling O1
→ Resolutions: 1k, 2k

​
Aspect Ratio Support

Most models support standard aspect ratios:
  • Square: 1:1
  • Landscape: 16:9, 4:3, 21:9
  • Portrait: 9:16, 3:4
  • Cinema: 3:2, 2:3

​
Advanced Features

Midjourney v7 offers unique creative controls: 
{
  "speed": "relaxed" | "fast" | "turbo",
  "variety": 0-100 (step: 5),
  "stylization": 0-1000,
  "weirdness": 0-3000
}
Ideogram v3 provides rendering modes: 
{
  "render_speed": "Turbo" | "Balanced" | "Quality",
  "style": "Auto" | "General" | "Realistic" | "Design"
}
1

For Beginners: Flux 2 Flex

Excellent balance of quality, speed, and ease of use. Supports resolution presets (1k/2k) and common aspect ratios.
2

For Photorealism: Google Imagen4

Produces highly realistic images quickly. Great for landscapes, products, and portraits.
3

For Artistic Work: Midjourney v7

Unmatched stylization options. Adjust variety, stylization, and weirdness for unique artistic styles.
4

For Iteration: Flux Schnell

Fastest generation time. Perfect for testing prompts and concepts before switching to quality models.
5

For Text Rendering: Ideogram v3

Best-in-class text generation. Ideal for logos, posters, and designs with readable text.
6

For Production: Flux 2 Pro or Imagen4 Ultra

Maximum quality for final deliverables. Worth the extra generation time.

​
Model Selection Workflow

​
Batch Generation

Many models support generating multiple images in a single request via the num_images parameter (1-4 images). Each image counts as a separate API charge.
Models supporting batch generation:
  • Flux series (most variants)
  • Google Imagen4 (except Ultra)
  • Ideogram v3
  • Bytedance Seedream v4
  • Qwen Image
  • Kling O1 (up to 9 images)

​
Advanced: LoRA Models

Flux Dev Lora allows custom style training: 
{
  "model_id": [
    {
      "model": "civitai:119351@317153",
      "weight": 1.0
    }
  ]
}
  • Find LoRA models on Civitai
  • Apply up to 4 LoRA models simultaneously
  • Weight range: 0-4 (default: 1)
  • Format: civitai:model_id@version_id
LoRA models require specific prompt keywords to activate. Check the Civitai model page for trigger words.

​
Next Steps

API Configuration

Set up your Muapi.ai API key

Cinema Studio

Master cinematic camera controls

Build docs developers (and LLMs) love

Get started for freeTalk to us