Owned end to end

Generation & Retrieval Studio

Everything on this page is generated, owned and measured. Images from my own GPU, sound synthesized live in your browser, retrieval over my own corpus. No API, no cloud, no claim without a number.

Sustainable by measurement, not by claim

Greenwashing is a claim without a number. Here are the numbers, measured on this very page and on the pipeline that produced it. Durable means: local, deterministic, reproducible, owned.

0API calls made by this page
0cloud GPUs used (all local CUDA)
-total weight of the owned catalogue
~10sper image on a 4 GB GPU, cost 0
Ownership

I own the full generate-and-retrieve stack

Not a wrapper around someone else's API. I build and own the generation, the retrieval and the governance across image, audio and video. Every asset carries its provenance; everything runs on my own GPU or on-prem.

Image

Generated on my own GPU (Stable Diffusion + LCM), free, deterministic, provenance baked in. Retrieval is BM25, on-prem, zero egress.

Audio

Synthesized live in your browser (Web Audio), nothing sampled, nothing licensed. The waveform and the scat riff are generated on the spot.

Video

Editable, governed storyboards the agent authors, scenes and on-screen data source-tagged. No stock, no third-party clip.

On the DarwinX theme I generate evolution, animals, nature and solar imagery on my own GPU. Where a real photo is needed, I source only public-domain and CC0 collections (Openverse, NASA, Smithsonian Open Access, Biodiversity Heritage Library), clearly attributed to their owner. What I claim to own, I generated.

Free generation

Free generation, owned end to end

Generated on my own NVIDIA GPU with Stable Diffusion + LCM (4-6 steps). Deterministic seeds: the same prompt and seed always reproduce the same pixels. License CreativeML OpenRAIL-M: the outputs are mine. Provenance is printed under each image, the anti-greenwashing rule applied to art.

DarwinX generated media reel
A reel of my own GPU-generated images (crossfade animation), all owned, provenance kept, built locally at zero cost.
Generate your own

Make your own image, free, on a shared GPU

You can generate your own image right in your browser, at no cost. It runs on a free cloud GPU (Colab or Kaggle) through a small notebook I built, Stable Diffusion 1.5 + LCM. Deterministic seeds, and you own your output (OpenRAIL-M). Non-commercial research demo only.

  1. Open 04_darwinx_public_gen_gradio.ipynb on Google Colab, set Runtime to GPU, run all cells.
  2. It prints a public gradio.live URL where anyone generates in the browser.
  3. Paste that URL below to embed your generator right here.

Free cloud GPU sessions are temporary and non-commercial. When the notebook stops, the URL expires, just re-run it. No key, no McLeuker resource, no stored data.

Generate art in your browser

Seeded and deterministic, like the GPU pipeline: same seed, same art. Generated on your device, zero network, you own the output.

Download seed 42
copyrightgenerated here, you own it
network calls0

Sound lab - waves, measured

A living waveform, synthesized live. Nothing sampled, nothing licensed, no claim: the measures print themselves below while it plays.

sample rate-
base frequency-
live RMS level-
copyrightsynthesized, nothing sampled
greenwashing check0 API · 0 egress · measured
i18n + l10n

Localization, live

Switch the language in the bottom-right widget (EN / FR / ع). The interface translates (i18n) and every field below re-localizes (l10n): numbers, currency, dates, direction. Arabic flips the page to RTL.

FieldLocalized value
Cost per generated image-
Catalogue images on my public CDN (production)-
Inference cost cut by KV-cache-
Today-