LMSupply.Captioner 0.111.0

LMSupply.Captioner

Local image captioning for .NET with automatic model downloading.

Features

• Zero-config: Models download automatically from HuggingFace
• GPU Acceleration: CUDA, CoreML (macOS)
• Cross-platform: Windows, Linux, macOS
• Simple API: Just 2 lines of code to get started

Quick Start

using LMSupply.Captioner;

// Load the default captioning model
var captioner = await LocalCaptioner.LoadAsync("default");

// Generate a caption
var result = await captioner.CaptionAsync("photo.jpg");
Console.WriteLine(result.Caption);
// Output: "A cat sitting on a windowsill looking outside"

Available Models

quality is published in quantized variants; a load takes the one the hardware tier picks, or the one you name with QuantizationHint or a qualifier ("quality:fp16"). LocalCaptioner.GetDownloadSizeBytesAsync answers what a load will download, for a consent screen.

using LMSupply.Captioner;

long bytes = await LocalCaptioner.GetDownloadSizeBytesAsync("quality");

await using var captioner = await LocalCaptioner.LoadAsync("quality",
    new CaptionerOptions { Detail = CaptionDetail.Detailed, MaxLength = 100 });
var result = await captioner.CaptionAsync("photo.jpg");

Detail (Brief · Detailed · Paragraph) is read by quality; default captions at one level and refuses any other value at load. Prompt (caption prefix) is read by default; quality refuses it — its prompt selects the task.

Advanced Usage

using LMSupply;   // ExecutionProvider

// Custom options
var options = new CaptionerOptions
{
    MaxLength = 50,
    Provider = ExecutionProvider.Cuda
};

var captioner = await LocalCaptioner.LoadAsync("default", options);
var result = await captioner.CaptionAsync("image.jpg");

Console.WriteLine($"Caption: {result.Caption}");
Console.WriteLine($"Confidence: {result.Confidence:P1}");

GPU Acceleration

Do not add ONNX Runtime packages (Microsoft.ML.OnnxRuntime*): LMSupply provisions the runtime itself, and a second copy conflicts with it. ExecutionProvider.Auto uses CUDA when the CUDA 12 runtime and cuDNN 9 are installed on the machine, CoreML on macOS, and the CPU otherwise. On Windows with an AMD or Intel GPU, ONNX sessions run on the CPU (DirectML was removed in 0.67.0). See GPU acceleration.