VibeSeq v0.1.2

VibeSeq v0.1.2

This unsigned release provides the VibeSeq desktop runtime, local persistence, arrangement editing, playback, and export.

Highlights

• Shift each Audio clip by up to one octave without changing its placement duration, and stretch time independently from 0.125x to 2x using Signalsmith Stretch.
• Record live input into an armed Audio track while listening to the Arrangement, with automatic browser input/output latency compensation and a manual trim control.
• Keep multi-pass loop recordings as one unfolded linear take for full editing, with REC N×, pitch, and stretch badges directly on the clip.
• Preserve raw recorded WAV media and apply latency correction as a non-destructive source offset. Waveform cache repair prevents uneven recording lengths from blocking durable saves.

It embeds the Studio, Stable Audio 3 execution source, a frozen portable CPU runtime, and verified first-run installers, so Node.js, Python, and uv do not need to be preinstalled. Model weights and the optional Windows CUDA runtime are not included in the executable.

On every launch VibeSeq checks the current platform's exact Stable Audio files. If VibeSeq Data is absent, empty, incomplete, or missing a required file, the Inference readiness panel offers a resumable SHA-256-verified download after the model terms are accepted:

• macOS ARM64 MLX bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-c2949a-macos-arm64
• Windows x64 CPU TFLite bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-c2949a-windows-x64
• Windows NVIDIA Ampere-or-newer bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-27b5a21-windows-x64-fa2. The app installs an isolated managed Python 3.12 runtime, official PyTorch 2.7.1 CUDA 12.6 wheels, and the digest-pinned community FlashAttention 2.8.3 wheel, runs an actual FlashAttention kernel on the detected GPU, then downloads the exact release checkpoint with the same one-click flow as CPU-only systems.
• Linux x64 TFLite bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-c2949a-linux-x64
• MuScriptor Medium, shared by every desktop platform: https://github.com/acidsound/VibeSeq/releases/tag/muscriptor-medium-f322369

Only the Stable Audio model format selected for the running hardware is downloaded. FlashAttention-compatible NVIDIA hardware must use the CUDA route and never silently downgrades to CPU. CPU-only Windows continues to use the public TFLite route. macOS and Linux store models below ~/VibeSeq Data/models; Windows stores models and its optional runtime below VibeSeq Data beside the installed VibeSeq.exe.

MuScriptor Medium is now available through the same one-click flow on macOS ARM64, Windows x64, and Linux x64. The app downloads the unchanged, digest-pinned public Release assets after the user accepts CC BY-NC 4.0 and the upstream conditions covering noncommercial use, rights in source audio, warranty disclaimer, and indemnification. The installed snapshot includes attribution, the exact upstream source revision, and license links. On NVIDIA Windows systems the same click also installs or repairs the isolated MuScriptor CUDA runtime; system Python is never modified.

Stable Audio and MuScriptor share the base Python/CUDA PyTorch environment and one persistent managed CUDA service on Windows, but install and verify as independent runtime profiles. Missing or broken MuScriptor packages never block Stable Audio readiness or generation. Stable Audio is loaded once and retains VRAM priority across generations; MuScriptor remains GPU-resident only when measured VRAM headroom allows it and otherwise stays on CPU without evicting Stable Audio. CUDA generation uses the official 8-step schedule and exact requested latent duration instead of generating an extra six seconds. FlashAttention 2 remains specific to Stable Audio, and the service never opens a console window.

Windows runtime installation clears and refreshes only the pinned FlashAttention wheel cache before synchronization, then automatically retries with an isolated temporary cache if interrupted installation metadata is incomplete. App shutdown terminates the complete sidecar, CUDA Python, and runtime-installer process trees; the persistent CUDA service also watches its parent process so it cannot remain orphaned and block a later Setup update.

Existing Windows users do not need to uninstall first. Close VibeSeq and run the 0.1.2 Setup over the same installation directory. The installer replaces the application in place and preserves the existing VibeSeq Data directory beside the executable.

Powered by Stability AI.

Windows is distributed as a current-user assisted NSIS installer so extraction progress is visible and VibeSeq Data remains beside the installed executable. The installed app then shows Local data, Audio engine, Health check, and Studio startup stages while its local service starts. Windows SmartScreen and macOS Gatekeeper may warn because code-signing identities are not configured yet. On Linux, run chmod +x VibeSeq-*.AppImage before launching. Verify the downloaded asset with SHA256SUMS.txt before running it.

Running the unsigned macOS build

1. Verify the downloaded DMG or ZIP against SHA256SUMS.txt.
2. Copy VibeSeq.app to /Applications.
3. Run:

xattr -dr com.apple.quarantine "/Applications/VibeSeq.app"

4. Launch VibeSeq again from /Applications.

VibeSeq v0.1.1

VibeSeq 0.1.1

This unsigned release provides the VibeSeq desktop runtime, local persistence, arrangement editing, playback, and export.

It embeds the Studio, Stable Audio 3 execution source, a frozen portable CPU runtime, and verified first-run installers, so Node.js, Python, and uv do not need to be preinstalled. Model weights and the optional Windows CUDA runtime are not included in the executable.

On every launch VibeSeq checks the current platform's exact Stable Audio files. If VibeSeq Data is absent, empty, incomplete, or missing a required file, the Inference readiness panel offers a resumable SHA-256-verified download after the model terms are accepted:

• macOS ARM64 MLX bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-c2949a-macos-arm64
• Windows x64 CPU TFLite bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-c2949a-windows-x64
• Windows NVIDIA Ampere-or-newer bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-27b5a21-windows-x64-fa2. The app installs an isolated managed Python 3.12 runtime, official PyTorch 2.7.1 CUDA 12.6 wheels, and the digest-pinned community FlashAttention 2.8.3 wheel, runs an actual FlashAttention kernel on the detected GPU, then downloads the exact release checkpoint with the same one-click flow as CPU-only systems.
• Linux x64 TFLite bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-c2949a-linux-x64
• MuScriptor Medium, shared by every desktop platform: https://github.com/acidsound/VibeSeq/releases/tag/muscriptor-medium-f322369

Only the Stable Audio model format selected for the running hardware is downloaded. FlashAttention-compatible NVIDIA hardware must use the CUDA route and never silently downgrades to CPU. CPU-only Windows continues to use the public TFLite route. macOS and Linux store models below ~/VibeSeq Data/models; Windows stores models and its optional runtime below VibeSeq Data beside the installed VibeSeq.exe.

MuScriptor Medium is now available through the same one-click flow on macOS ARM64, Windows x64, and Linux x64. The app downloads the unchanged, digest-pinned public Release assets after the user accepts CC BY-NC 4.0 and the upstream conditions covering noncommercial use, rights in source audio, warranty disclaimer, and indemnification. The installed snapshot includes attribution, the exact upstream source revision, and license links. On NVIDIA Windows systems the same click also installs or repairs the isolated MuScriptor CUDA runtime; system Python is never modified.

Stable Audio and MuScriptor share the base Python/CUDA PyTorch environment and one persistent managed CUDA service on Windows, but install and verify as independent runtime profiles. Missing or broken MuScriptor packages never block Stable Audio readiness or generation. Stable Audio is loaded once and retains VRAM priority across generations; MuScriptor remains GPU-resident only when measured VRAM headroom allows it and otherwise stays on CPU without evicting Stable Audio. CUDA generation uses the official 8-step schedule and exact requested latent duration instead of generating an extra six seconds. FlashAttention 2 remains specific to Stable Audio, and the service never opens a console window.

Windows runtime installation clears and refreshes only the pinned FlashAttention wheel cache before synchronization, then automatically retries with an isolated temporary cache if interrupted installation metadata is incomplete. App shutdown terminates the complete sidecar, CUDA Python, and runtime-installer process trees; the persistent CUDA service also watches its parent process so it cannot remain orphaned and block a later Setup update.

Existing Windows alpha users do not need to uninstall first. Close VibeSeq and run the 0.1.1 Setup over the same installation directory. The installer replaces the application in place and preserves the existing VibeSeq Data directory beside the executable.

Powered by Stability AI.

Windows is distributed as a current-user assisted NSIS installer so extraction progress is visible and VibeSeq Data remains beside the installed executable. The installed app then shows Local data, Audio engine, Health check, and Studio startup stages while its local service starts. Windows SmartScreen and macOS Gatekeeper may warn because code-signing identities are not configured yet. On Linux, run chmod +x VibeSeq-*.AppImage before launching. Verify the downloaded asset with SHA256SUMS.txt before running it.

Running the unsigned macOS build

Until VibeSeq is signed and notarized with an Apple Developer ID, macOS may report that the downloaded app is damaged or cannot be opened. After verifying the official Release asset with SHA256SUMS.txt, copy VibeSeq.app to /Applications, open Terminal, and run:

xattr -dr com.apple.quarantine "/Applications/VibeSeq.app"

Then open VibeSeq again by double-clicking its icon in Applications. Run this command only for the VibeSeq app downloaded from this official Release.

VibeSeq v0.1.0-alpha.18

Desktop runtime validation prerelease

This unsigned prerelease is for clean-machine validation of the VibeSeq desktop runtime, local persistence, arrangement editing, playback, and export.

It embeds the Studio, Stable Audio 3 execution source, a frozen portable CPU runtime, and verified first-run installers, so Node.js, Python, and uv do not need to be preinstalled. Model weights and the optional Windows CUDA runtime are not included in the executable.

On every launch VibeSeq checks the current platform's exact Stable Audio files. If VibeSeq Data is absent, empty, incomplete, or missing a required file, the Inference readiness panel offers a resumable SHA-256-verified download after the model terms are accepted:

• macOS ARM64 MLX bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-c2949a-macos-arm64
• Windows x64 CPU TFLite bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-c2949a-windows-x64
• Windows NVIDIA Ampere-or-newer bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-27b5a21-windows-x64-fa2. The app installs an isolated managed Python 3.12 runtime, official PyTorch 2.7.1 CUDA 12.6 wheels, and the digest-pinned community FlashAttention 2.8.3 wheel, runs an actual FlashAttention kernel on the detected GPU, then downloads the exact release checkpoint with the same one-click flow as CPU-only systems.
• Linux x64 TFLite bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-c2949a-linux-x64
• MuScriptor Medium: https://huggingface.co/MuScriptor/muscriptor-medium

Only the Stable Audio model format selected for the running hardware is downloaded. FlashAttention-compatible NVIDIA hardware must use the CUDA route and never silently downgrades to CPU. CPU-only Windows continues to use the public TFLite route. macOS and Linux store models below ~/VibeSeq Data/models; Windows stores models and its optional runtime below VibeSeq Data beside the installed VibeSeq.exe.

MuScriptor remains gated and is not redistributed. Each user accepts its Hugging Face conditions, downloads the pinned config.json and model.safetensors in their browser, and places them directly under SAVE CACHE UNDER. That exact revision folder opens in Finder, File Explorer, or the Linux file manager. VibeSeq does not download, select, or move these files or request or store a Hugging Face token; Verify files in cache only validates the files already in that folder.

Stable Audio and MuScriptor share the base Python/CUDA PyTorch environment and one persistent managed CUDA service on Windows, but install and verify as independent runtime profiles. Missing or broken MuScriptor packages never block Stable Audio readiness or generation. Stable Audio is loaded once and retains VRAM priority across generations; MuScriptor remains GPU-resident only when measured VRAM headroom allows it and otherwise stays on CPU without evicting Stable Audio. CUDA generation uses the official 8-step schedule and exact requested latent duration instead of generating an extra six seconds. FlashAttention 2 remains specific to Stable Audio, and the service never opens a console window.

Windows runtime installation clears and refreshes only the pinned FlashAttention wheel cache before synchronization, then automatically retries with an isolated temporary cache if interrupted installation metadata is incomplete. App shutdown terminates the complete sidecar, CUDA Python, and runtime-installer process trees; the persistent CUDA service also watches its parent process so it cannot remain orphaned and block a later Setup update.

Powered by Stability AI.

Windows is distributed as a current-user assisted NSIS installer so extraction progress is visible and VibeSeq Data remains beside the installed executable. The installed app then shows Local data, Audio engine, Health check, and Studio startup stages while its local service starts. Windows SmartScreen and macOS Gatekeeper may warn because code-signing identities are not configured yet. On Linux, run chmod +x VibeSeq-*.AppImage before launching. Verify the downloaded asset with SHA256SUMS.txt before running it.

macOS installation

This prerelease is not signed or notarized with an Apple Developer ID, so macOS may report that VibeSeq is damaged.

1. Open the DMG and drag VibeSeq.app into Applications.
2. Open Terminal and run:

xattr -dr com.apple.quarantine "/Applications/VibeSeq.app"

Double-click VibeSeq in Applications.
This command only removes the browser-added quarantine attribute from VibeSeq. It does not disable Gatekeeper system-wide and does not require sudo.
Only run it after downloading VibeSeq from the official GitHub release and verifying SHA256SUMS.txt.

VibeSeq v0.1.0-alpha.11

Desktop runtime validation prerelease

This unsigned prerelease is for clean-machine validation of the VibeSeq desktop runtime, local persistence, arrangement editing, playback, and export.

It embeds the Studio, Stable Audio 3 execution source, frozen real-model runtime, and the verified first-run model installer, so Node.js, Python, and uv are not required. Model weights are not included in the executable.

On every launch VibeSeq checks the current platform's exact Stable Audio files. If VibeSeq Data is absent, empty, incomplete, or missing a required file, the Inference readiness panel offers a resumable SHA-256-verified download after the model terms are accepted:

• macOS ARM64 MLX bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-c2949a-macos-arm64
• Windows x64 TFLite bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-c2949a-windows-x64
• Linux x64 TFLite bundle: https://github.com/acidsound/VibeSeq/releases/tag/stable-audio-3-c2949a-linux-x64
• MuScriptor Medium: https://huggingface.co/MuScriptor/muscriptor-medium

Only the Stable Audio model format for the running OS is downloaded. macOS and Linux store it below ~/VibeSeq Data/models; Windows stores it below VibeSeq Data/models beside the installed VibeSeq.exe.

MuScriptor remains gated and is not redistributed. Each user accepts its Hugging Face conditions, downloads the pinned config.json and model.safetensors in their browser, and places them directly under SAVE CACHE UNDER. That exact revision folder opens in Finder, File Explorer, or the Linux file manager. VibeSeq does not download, select, or move these files or request or store a Hugging Face token; Verify files in cache only validates the files already in that folder.

Powered by Stability AI.

Windows is distributed as a current-user assisted NSIS installer so extraction progress is visible and VibeSeq Data remains beside the installed executable. The installed app then shows Local data, Audio engine, Health check, and Studio startup stages while its local service starts. Windows SmartScreen and macOS Gatekeeper may warn because code-signing identities are not configured yet. On Linux, run chmod +x VibeSeq-*.AppImage before launching. Verify the downloaded asset with SHA256SUMS.txt before running it.

MuScriptor Medium · all desktop platforms · f322369

Digest-pinned MuScriptor Medium model bundle for VibeSeq one-click installation on macOS ARM64, Windows x64, and Linux x64. Developed by Mirelo x Kyutai. Weights are redistributed without modification under CC BY-NC 4.0 and remain noncommercial-only. See MuScriptor-NOTICE.txt.

Stable Audio 3 Medium · Windows x64 CUDA/FlashAttention 2 · 27b5a21

Windows x64 first-run model bundle for the NVIDIA CUDA/FlashAttention 2 route. These are the exact unmodified Stable Audio 3 Medium and T5Gemma files at the pinned revision, split only to satisfy the GitHub Release 2 GiB per-file limit. Installation requires explicit acceptance of the Stability AI Community License and Gemma Terms of Use inside VibeSeq. Powered by Stability AI.

Stable Audio 3 Medium · Linux x64 TFLite · c2949a

Linux x64 first-run model bundle. The model files are the official optimized TFLite medium weights used by the CPU/LiteRT runtime. Powered by Stability AI.

Stable Audio 3 Medium · Windows x64 TFLite · c2949a

Platform-specific Stable Audio 3 Medium optimized TFLite weights for VibeSeq on Windows x64. Includes the Stability AI Community License, Gemma Terms of Use, and required notice. The app verifies every asset and model file against the bundled SHA-256 manifest before activation. Powered by Stability AI.

Stable Audio 3 Medium · macOS ARM64 MLX · c2949a

Platform-specific Stable Audio 3 Medium optimized weights for VibeSeq on Apple Silicon. Includes the Stability AI Community License, Gemma Terms of Use, and required notice. The app verifies every asset and reconstructed model file against the bundled SHA-256 manifest before activation. Powered by Stability AI.