How Inverity Works
From classical compression to cutting-edge neural codecs, Inverity delivers the smallest possible files at your target quality level, automatically integrated with your CMS.
The process
The Inverity Workflow
Five steps from upload to deployment.
Step 1 — Upload or Connect
Upload images directly or connect your CMS file manager via OAuth. Inverity scans your existing library to identify optimization opportunities.
Step 2 — Neural Analysis
Deep learning models analyze each image for content type, complexity, and optimal compression strategy. Face detection and saliency mapping identify important regions.
Step 3 — Intelligent Compression
Based on your selected profile, Inverity applies the right backend, from classical codecs to cutting-edge neural compression, to maximize savings while preserving quality.
Step 4 — Quality Verification
Every optimized asset is verified using SSIM, PSNR, and MAE metrics. Images that do not meet quality thresholds are flagged or rejected automatically.
Step 5 — Deploy or Review
Approved assets sync back to your CMS automatically, or enter an approval workflow for review. Version history and rollback are always available.
Neural Compression Backends
Inverity Enterprise unlocks the most advanced compression technology available: deep learning models that outperform classical codecs at every quality level.
End-to-End Neural Pipeline (Enterprise only)
The most advanced compression Inverity offers. Every stage — encoding, resizing, format conversion, delivery variant generation — runs through neural models trained on millions of images. Smallest possible file sizes at target quality. Perceptual optimization focused on what humans notice. Adaptive: learns optimal settings per image. Delivery variants generated with neural resizers.
Neural Compression (Pro & Enterprise)
AI-powered compression at quality indistinguishable from the original. Every encoded variant is verified per-image for PSNR and SSIM; if the neural quality SLO is not met for a given image, we transparently fall back to JPEG-XL lossless. The output is up to 9× smaller than standard lossless on UI/marketing content — and never worse than the best classical codec. Per-image quality verification on every variant. Automatic JPEG-XL lossless fallback when SLO is not met. Model version pinned per tenant for reproducibility.
Under the Hood
How neural compression actually works.
Neural Architecture
Inverity neural backends use learned entropy models and deep convolutional architectures trained on diverse image datasets. Models replace traditional statistical entropy coders with neural networks that predict probability distributions more accurately.
Model Versioning
Every neural backend is version-pinned per tenant. Model manifests are loaded at task start and cached. This ensures reproducibility and allows rollback to previous model versions if needed.
Benchmark Freshness
Neural backends are continuously benchmarked against quality datasets. Benchmark age and recertification status are tracked. Backends exceeding freshness thresholds are flagged for recertification before production rollout.
Learned Entropy Coding
An intermediate step between classical and full neural compression. Replaces the entropy coder in classical pipelines with an AI-trained model, producing smaller bitstreams without changing the quantization stage.
Quality Verification
Every optimized asset is validated before deployment.
SSIM — Structural Similarity Index
Measures perceived structural change. Closely correlates with human perception of quality loss.
PSNR — Peak Signal-to-Noise Ratio
Quantifies pixel-level difference between original and optimized. Higher values indicate better fidelity.
MAE — Mean Absolute Error
Measures average pixel deviation. Used to detect compression artifacts and edge degradation.
Quality Thresholds: each optimization profile defines pass/warn/fail thresholds for these metrics. Assets that fail QA checks are automatically flagged and never deployed without explicit approval.
Seamless CMS Integration
Inverity integrates with your existing content management workflow.
OAuth 2.0 Connection
Connect your CMS via secure OAuth 2.0. Inverity only requests file manager permissions, no access to contacts, forms, or other sensitive data. Revocable at any time from your CMS settings.
Workflow Automation
Inverity integrates with your CMS workflow automation. Every image that hits your trigger point gets optimized automatically, no manual upload step required. Signed webhook verification ensures secure communication.
Bidirectional Sync
Optimized assets sync back to your CMS automatically. Choose between parallel versions (keeps original) or in-place replacement. Token refresh and fallback strategies ensure reliability.
Cache Invalidation
When an optimized version replaces an existing asset, Inverity purges downstream CDN caches automatically. Your end users always serve the latest variant — no manual purge step, no stale-image surprises after a publish.
Ready to try neural compression?
Join the Waitlist