planned project / documentation
Ellastic
An experimental media manipulation framework built around a simple idea: digital media is ultimately data.
Ellastic is intended to let people inspect, transform, layer and intentionally corrupt the bytes, structures, signals and encoded information beneath audio, images, video and other files.
This is a description of the intended project, based on its feature plan. Ellastic is unfinished and is not currently runnable.
01 / foundations
One abstraction for many kinds of media
A shared media abstraction is planned so operations can work with raw bytes, metadata, streams, frames, samples, channels and format information without needing to know how a file was loaded. Format detection is intended to use magic bytes, headers, MIME types and extensions, while unknown files remain usable as raw data.
Raw binary mode
Planned access to offsets, hexadecimal and binary representations, decimal values, byte ranges, file size and entropy, even when a file is not recognized.
Preserve the source
Processing is designed to be non-destructive by default: input flows through an operation into a new output. Random operations should accept explicit seeds so results can be reproduced.
02 / composition
Transformation as a pipeline
The pipeline is the center of the concept. Operations are intended to be layered, structured and serializable, with each operation receiving the previous output.
Parameters and serialization
Operation parameters, input and output configuration, and random seeds should be saved in executable pipeline files.
Preview
Reduced previews are planned for images, audio, video and raw binary changes before an export is made.
03-04 / deliberate change
Databending, corruption and controlled destruction
These are primary purposes of Ellastic. Planned byte mutation includes replacement, insertion, removal, duplication, swapping, reversal, rotation, shuffling, repetition, truncation and expansion. Byte arithmetic, bit operations, XOR modes and configurable block operations extend that control.
Corruption engine
Random, burst, sparse, progressive and region-based corruption should work with probability, strength, distributions, seeds and masks.
Regions
Operations may target byte ranges, percentages, random or repeating regions, and conditional matches such as a byte value, pixel brightness or sample amplitude.
05-10 / media layers
Signals, audio, images, video and containers
The plan moves between decoded and encoded layers rather than treating every format as the same. Signals may be extracted into samples and channels for amplitude operations, mathematics, oscillation, modulation, folding and sample manipulation.
Audio
Initial import targets are WAV, AIFF, FLAC, OGG and MP3. Planned work includes export settings, channel operations, channel-specific corruption, sample-rate and bit-depth manipulation.
Images and image databending
PNG, JPEG, BMP, TIFF, WebP and GIF are listed as initial targets. Pixel and channel work sits alongside header preservation, payload corruption, compression-aware operations and re-encode loops.
Video and containers
Video plans include streams, frames, keyframes, motion data and codec-aware datamoshing. Containers should expose tracks, metadata, packets and independent stream extraction or replacement.
11-16 / inspection
Make the changes visible
Ellastic is also intended to explain what happened. Planned tools include procedural data generation, a hex viewer, byte diffs, entropy and histogram analysis, waveforms, spectra, spectrograms and image differences.
Randomness and reversibility
Seeded distributions, random operation selection and generated parameters should be explicit. Operations should identify whether they are reversible, partially reversible or irreversible.
Operation metadata
Each operation should describe its name, category, determinism, reversibility, lossiness and supported media types.
17-21 / future tools
A framework that can grow
Third-party transformations, codecs, analysis tools, generators and formats are planned through a plugin API. A CLI is intended for scripting, inspection, diffs and batch processing, while a graphical interface may eventually provide a pipeline editor, live previews, an operation browser and before/after comparison.
Presets, recursive and parallel processing, filename templates and watch mode are also part of the planned automation layer.
22-29 / direction
Roadmap, principles and the long view
Graceful failure, salvage, streaming, memory mapping, incremental processing, provenance logs and file-protection guardrails are planned alongside progressive format support. The roadmap below is aspirational, not a list of shipped functionality.
Project architecture exists; media abstraction, loading, writing, detection, operations, pipelines, errors and tests remain planned.
Byte and bit operations, arithmetic, random corruption and seeded randomness.
PNG and BMP decoding/encoding, pixels, channels, slices, scanlines, diffs and histograms.
WAV, samples, amplitude, signal mathematics, oscillators, noise, modulation and visualization.
JPEG payloads, compression-aware corruption, re-encoding, metadata and containers.
Video datamoshing, batch workflows, interactive tools and a plugin ecosystem.
The long-term goal is to move through a workflow from file to container, stream, encoded data, decoded data, signal, pixels or samples and mathematical representation, then back again. The process should remain controllable, inspectable and reproducible.