For designers and builders using ChatGPT images

Turn hazy ChatGPT images into crisp PNG and SVG files.

HazeFix is designed to reduce sepia haze and diffusion noise, upscale by 4× and trace flat-colour artwork to SVG. Add an image, choose a preset and review the result.

Fix an image View API reference

10 images a month on Free · no credit card · full output quality, no watermark

ChatGPT character illustration, before and after HazeFix

BEFOREAFTER

Source palette→reduced paletteSource size→4× PNG + SVG

sample pipeline output — drag the handle

Studio

Add an image. Choose your finish.

Pick a preset and add a PNG, JPEG or WebP image. A free account includes 10 images a month at full output quality.

Drop your ChatGPT image

Click to browse, or paste with ⌘V straight from ChatGPT

PNG · JPEG · WebP · up to 24 MB

ADVANCED

Palette24 colours

The pipeline

Five stages, tuned to one failure mode.

Generic upscalers optimise for photorealism, while generic vectorisers assume clean input. HazeFix combines five stages for flat-colour AI artwork affected by haze and diffusion noise.

De-haze

A fitted white-balance profile targets sepia haze and nudges the tonal range towards cleaner blacks and whites.

warmer cast → more neutral

Flatten

Edge-preserving filtering reduces diffusion sampling noise in regions intended to read as one flat colour.

regional noise → reduced

Upscale 4×

An illustration-tuned Real-ESRGAN pass, configured for clean fills and crisp line art rather than photo texture.

source size → 4× output

Reduce palette

k-means in CIELAB groups similar colours into a smaller palette; boundary sharpness depends on the source and preset.

source colours → selected palette

Trace

vtracer runs after flattening and palette reduction to produce an editable SVG from the processed source.

PNG → true SVG

Plain-language facts

What HazeFix does — and what agents can rely on.

What is HazeFix?

A web image-processing tool for flat-colour artwork. It combines colour correction, edge-preserving filtering, optional 4× upscaling, palette reduction and SVG tracing. Results depend on the source image and preset, so review each output before use.

What files are supported?

The Studio and API accept PNG, JPEG and WebP files up to 24 MB. Completed jobs can return a processed PNG and, when requested, a traced SVG.

How can software use it?

The documented REST API creates a job, reports its status and serves its output files. It uses HazeFix bearer API keys; HazeFix does not currently publish OAuth, MCP, WebMCP or agent-registration endpoints.

For builders

The same pipeline, one POST away.

One POST, one poll

Create a job, poll until done, download PNG and SVG. The whole surface is three endpoints.

Metrics in the response

Every job returns before/after colour counts, histogram endpoints and stage timings for monitoring or automated checks.

Built for batch

Submit jobs from scripts or batch workflows. API pricing is $0.03 per completed image, with volume pricing at 10,000+ images a month.

View API reference

CREATE A JOB

curl -s https://api.hazefix.com/api/v1/jobs \  -H "Authorization: Bearer $HAZEFIX_API_KEY" \  -F "image=@./chatgpt-logo.png" \  -F "preset=logo"# → 202 Accepted{  "id": "9f2b1c4e-…",  "status": "processing",  "preset": "logo",  "options": { "k": 12, "upscale": true, "svg": true }}