MidJourneyV7Model<T>

Models & Types · AiDotNet.Diffusion.TextToImage

MidJourney V7-style model architecture for artistic text-to-image generation.

Inheritance: DiffusionModelBase<T> → LatentDiffusionModelBase<T> → MidJourneyV7Model<T>

For Beginners

MidJourney V7 is renowned for producing the most visually stunning and artistic images in the AI generation space.

How MidJourney V7 works:

  1. Text is deeply interpreted by a proprietary language model for nuanced understanding
  2. A multi-scale MMDiT-X processes tokens at different resolution scales
  3. Aesthetic-aware training ensures consistently beautiful output
  4. A stylize parameter controls artistic interpretation vs literal prompt following

Key characteristics:

  • Multi-scale MMDiT-X for coherent generation across scales
  • Proprietary language model for deep prompt interpretation
  • Aesthetic-aware training with human preference data
  • Stylize parameter (0-1000) controlling artistic expression
  • Strong photorealistic and artistic capabilities
  • 16 latent channels

Advantages:

  • Best-in-class artistic and aesthetic quality
  • Exceptional at photorealism and creative composition
  • Strong prompt interpretation and creative expansion
  • Stylize parameter gives fine control over output style

Limitations:

  • API-only (not open-source, architecture details are proprietary)
  • Less precise prompt adherence when stylize is high
  • Requires internet connection and subscription

How It Works

Represents the MidJourney V7-style architecture known for highly artistic and photorealistic generation. Uses a multi-scale MMDiT-X architecture with aesthetic-aware training, enhanced prompt interpretation via a proprietary language model, and a stylize parameter for controlling the artistic vs photorealistic balance.

Technical specifications:

  • Architecture: Multi-scale MMDiT-X with aesthetic training
  • Hidden size: estimated ~4096, ~40+ layers
  • Stylize range: 0 (literal) to 1000 (highly artistic)
  • VAE: 16 latent channels
  • Default: 50 steps, guidance scale 5.0
  • Resolution: 1024x1024 default, aspect-ratio aware

Note: MidJourney is proprietary; this is a best-effort architectural representation based on publicly available information.

Example

// Create MidJourney V7
var mj7 = new MidJourneyV7Model<float>();
            
// Generate an artistic image
var image = mj7.GenerateFromText(
    prompt: "ethereal forest spirit emerging from morning mist, cinematic lighting",
    width: 1024,
    height: 1024,
    numInferenceSteps: 50,
    guidanceScale: 5.0,
    seed: 42);

Constructors

Constructor Summary
MidJourneyV7Model(NeuralNetworkArchitecture<T> architecture, DiffusionModelOptions<T> options, INoiseScheduler<T> scheduler, MMDiTXNoisePredictor<T> predictor, StandardVAE<T> vae, IConditioningModule<T> conditioner, int? seed)

Properties

Property Summary
IConditioningModule<T> Conditioner { get; }
INoisePredictor<T> NoisePredictor { get; }
IVAEModel<T> VAE { get; }
int LatentChannels { get; }

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()
Tensor<T> GenerateFromText(string prompt, string negativePrompt, int width, int height, int numInferenceSteps, double? guidanceScale, int? seed)
Tensor<T> ImageToImage(Tensor<T> inputImage, string prompt, string negativePrompt, double strength, int numInferenceSteps, double? guidanceScale, int? seed)

Fields

Field Summary
int DefaultHeight
int DefaultWidth