drizzymedia/synapse-vision

SynapseVLM Caption Evaluator

Drop multiple images, select a model, and compare generated captions to determine which Synapse vision model is best suited for captioning your image datasets.

Models (7–9B range, one resident at a time)

SynapseVLM is a native multimodal vision-language foundation model designed for image understanding, visual reasoning, and high-quality image caption generation.

Thinking mode is disabled by default for clean, consistent captions.

Hardware

Requires SynapseGPU (48 GB large inference slice).

Models are loaded on first use or when switching models (~1 minute). Caption generation typically completes within seconds per image after the model is loaded.