01 / Translate
Choose a mode, listen, then change a parameter. Use repeatable mappings to investigate the effect.
AN AUDIOVISUAL RESEARCH ENVIRONMENT
See an image.
Hear a possibility.
Turn images into sound, study what changes, and preserve the work as an archive you can listen to, compare and publish.
For artists, designers, sound makers and curious researchers. Begin with a sample, a library image, generated artwork or a shared gallery study. Explore freely, or try one of 12 guided experiments.
Unseen Scores is built by Crafted Tools. We bring image, sound and pattern into one research environment, giving creative practitioners a place to explore audiovisual mappings, preserve their experiments and share what they discover.
SPECIMEN 001 / COLOR FIELD
One image, four ways of listening.
Visual columns become pitched notes and timing. Six seconds, generated in your browser by the same engine as the workspace.
No account or upload needed for this sample. It does not use AI.
A practice, from first experiment to finished edition.
02 / Preserve
Save the sampled source, exact audio and settings. Add a question and observations; branch into another experiment.
03 / Share the work
Share a preserved study in the public gallery with credit, arrange a listening collection, assemble an edition, or prepare a waveform print.
Choose a research lens.
Spectral
Brightness and position become frequency bands.
Musical
Visual columns become pitched notes and timing.
Environmental
Image values shape layered, evolving tones.
Pattern
Visual structure becomes a rhythmic sequence.
Cipher and Reconstruction add exact sampled-image encoding experiments. Optional AI interpretation, live capture and further exploratory labs extend the practice.
Before you begin
What can I make?
Image-to-sound studies, reusable recipes and presets, comparison sets, listening collections, editions, printable books and PDF/SVG print proofs. Four core mappings interpret a 64 × 64 sample over 1–120 seconds. Their WAV exports use native 48 kHz / 24-bit audio, with optional stereo finishing. You can also generate abstract artwork and share selected studies in the public gallery.
Do I need an account?
Upload an image, choose a library image or generate one, then explore Spectral, Musical, Environmental or Pattern mode and download without signing in. Anyone can browse and adapt gallery studies. Accounts unlock saving, comparisons, collections, gallery publishing, editions, print tools, Cipher, Reconstruction and the additional research labs.
Can I generate and download artwork?
The image generator creates purely abstract, black-and-white artwork. Download a PNG with a 4K (4,096 px) or 6K (6,144 px) long edge, or choose one of the four sound mappings. These downloads enlarge the generated image without adding new detail; sound still uses a 64 × 64 sample. There are three attempts per browser per day and a shared site limit of 30. Failed attempts count; limits reset at midnight UTC.
How does the Study Gallery work?
From your saved study in one of the four core modes, publish a version with its source sample and rendered WAV preserved. Add a public credit and optional description, and confirm permission to share. It appears immediately with a public copy of the 64 × 64 source, saved sound, settings and provenance, not your private notebook. Others can adapt it with credit without changing the original. You can remove your entry, but public files and downloaded copies cannot be recalled.
What stays private?
New studies, recipes, presets and notebooks start private to you and app administrators. Uploaded images are processed locally; saved studies always keep a private 64 × 64 source sample, with optional WAV and manifest preservation. Library, generated and gallery images use public links. Generated results remain available in the same tab for up to 24 hours, not as saved library records. Editions are readable by signed-in app users; the gallery is public. Research bundles include saved notebook text.
Does every experiment use AI?
No. The four core mappings and guided experiments run deterministically in your browser. The image generator sends your prompt to an AI image service and returns a public image link. Optional AI Interpretation sends your selected image for analysis; AI features in the additional labs send the entered colors or data description when requested. Do not include confidential material in public prompts.
What is still experimental?
Cipher encodes a 128 × 128 RGBA sample into its own checked WAV format; Reconstruction recovers that sample and can compare it with a reference. This is not encryption and cannot recover an original image from arbitrary audio. Live Translation captures still images, not continuous video sonification. MP3/FLAC exports, full-resolution original preservation in studies, long-form recording and native book PDF generation remain future work. Enlarged artwork PNGs and printable HTML books are available now.
How do I help shape it?
Explore at your own pace, then use Feedback inside the app to share a problem or idea. A research question is optional, not a required task. Create an account to save what you discover; no mailing-list subscription is added automatically.