Instrument Plugin · VST3 / CLAP / Standalone

AkiBloom

Grow new playable timbres from your own samples through three organic operations — graft, hybridize, cultivate.

License GPL-3.0Platform Windows x64Format VST3 · CLAP · Standalone (WASAPI)Language Rust · nih-plug
Features

What it does

AkiBloom is an instrument plugin built on three organic operations: graft one seed's timbre toward another, hybridize it along metaphor directions (wet · grainy · breathing), and cultivate it with keep/discard feedback and a walkable timbre map. The interaction framework comes from the research project Botanica-OS, realized with no neural runtime required — pure real-time-safe DSP. Every grown timbre exports as WAV and can be used as a parent in the sibling effect AkiGraft.

🌱

Three organic operations

Graft pulls one seed's timbre toward another (α = 0 keeps the original sound); hybridize varies it along metaphor directions; cultivate evolves it with your feedback.

🎲

Deterministic hybridize

Pick a direction (wet / grainy / breathing), set the amount, reroll the seed — same seed, same variation, always.

🗺️

Cultivate + timbre map

A walker takes over the variation; audition each mutation on the keys, keep or discard, and the timbre map remembers every step — click any dot to roll back to it.

🎹

Playable

Click the on-screen keyboard (C3–C5) or plug in a MIDI keyboard; set the native pitch root (C4 = 60) and ±1 semitone fine tune.

💾

WAV export

Export the current timbre as WAV, ready to drop into the sibling AkiGraft effect as a parent — the two plugins form one closed workflow.

⚡

Real-time-safe DSP

No neural runtime required — pure DSP. Loads two of your own WAV seeds (16/24/32-bit, mono or stereo; longer than 10 s is trimmed).

Details

Controls

ControlWhat it does
源 Source A/BWhich seed is the played source; the other donates its timbre
生长量 Grow αHow far the source is pulled toward the donor (0 = original)
方向 / 变异量 / 种子Hybridize direction, depth, deterministic seed
培育 CultivateWalker takes over the variation; keep/discard + map rollback
根音 Root / 微调 TuneNative pitch note (C4 = 60) and ±1 semitone fine tune
输出 OutputOutput level
Quick Start

Grow your first timbre

  1. Load two of your own WAV samples as seeds (16/24/32-bit, mono or stereo; longer than 10 s is trimmed).
  2. Drag Grow α — the source seed is pulled toward the other seed's timbre; α = 0 keeps the original sound.
  3. Play the on-screen keyboard (C3–C5) or plug in a MIDI keyboard.
  4. Hybridize: pick a direction (wet / grainy / breathing), set the amount, reroll the seed — same seed, same variation.
  5. Cultivate: audition each mutation, keep or discard, roll back anywhere on the map — then export the timbre as WAV.
shell — build from source
cargo test --release
cargo xtask bundle akibloom --release   # VST3 + CLAP + standalone
python scripts/verify_render.py         # offline render self-test
乐器插件 · VST3 / CLAP / 独立运行

项目简介

AkiBloom 是一件建立在三种「有机化操作」上的乐器插件:把一颗种子的音色向另一颗嫁接拉拢,沿隐喻方向(湿润 · 颗粒 · 呼吸感)做杂交变异,再以保留/丢弃反馈与可游走的音色地图加以培育。交互框架来自研究项目 Botanica-OS 的产品化落地——无需神经网络运行时,纯实时安全 DSP。长好的音色可一键导出 WAV,直接当作姊妹效果器 AkiGraft 的母本。

🌱

三种有机化操作

嫁接把一颗种子的音色向另一颗拉拢(α = 0 保持原声);杂交沿隐喻方向做变异;培育以你的反馈引导演化。

🎲

确定性杂交

选方向(湿润 / 颗粒 / 呼吸感)、推变异量、换种子——同一种子永远得到同一个变异。

🗺️

培育与音色地图

演化交给 walker;在琴键上试听每个变异,👍保留或✗丢弃,地图记住每一步——点击任意圆点即可回溯。

🎹

可演奏

点击界面琴键(C3–C5)或接入 MIDI 键盘;设定本调根音(C4 = 60)与 ±1 半音微调。

💾

WAV 导出

把当前音色导出为 WAV,可直接拖入姊妹效果器 AkiGraft 当母本——两件插件构成一个完整闭环。

⚡

实时安全 DSP

无需神经网络运行时,纯 DSP。加载两颗你自己的 WAV 采样作种子(16/24/32bit,单声道或立体声;超 10 秒自动截取)。

明细

控制项

控制项作用
源 Source A/B决定演奏的源种子;另一种子贡献其音色
生长量 Grow α源向供体音色靠拢的程度(0 = 原声)
方向 / 变异量 / 种子杂交方向、深度与确定性种子
培育 Cultivate演化交给 walker;保留/丢弃 + 地图回溯
根音 Root / 微调 Tune本调音高(C4 = 60)与 ±1 半音微调
输出 Output输出电平
  1. 加载两颗你自己的 wav 采样作种子(16/24/32bit,单声道或立体声;超 10 秒自动截取)。
  2. 推生长量 α——源种子向另一种子的音色靠拢;α = 0 保持原声。
  3. 演奏:点击界面琴键(C3–C5)或接入 MIDI 键盘。
  4. 杂交:选方向(湿润/颗粒/呼吸感)、推变异量、换种子——同一种子永远得到同一个变异。
  5. 培育:试听每个变异,保留或丢弃,在地图上任意回溯——最后把音色导出为 WAV。
shell — build from source
cargo test --release
cargo xtask bundle akibloom --release   # VST3 + CLAP + standalone
python scripts/verify_render.py         # offline render self-test