Skip to content
SPECIFICTION
Get listed

StemForge

github.com

Free plan Open source WebWindowsmacOSLinuxSelf-hosted
Visit github.com
Spec score 8.2 #12 of 22 in AI Music Remixing Tools

What it is

StemForge is a locally run, browser-based audio workstation for separating stems, extracting MIDI, generating audio, composing songs, mixing, and exporting. Demucs and BS-Roformer can separate audio into vocal, drum, bass, other, guitar, and piano stems; spectral analysis informs engine and model recommendations. For MIDI extraction, StemForge uses BasicPitch on instrument stems and faster-whisper with pitch tracking on vocal melodies. Its Synth tab generates text-conditioned audio with Stable Audio Open for up to 600 seconds, with optional audio or MIDI conditioning. Compose uses AceStep for song generation and RVC for voice conversion, with 14 built-in voices and a HuggingFace model browser. Mix combines stems, MIDI-rendered tracks, synth outputs, and composed songs; exports include WAV, FLAC, MP3, and OGG, individually or in a ZIP. The project says audio remains on the user's computer and no internet is needed after initial model downloads. Setup requires Python 3.12, FFmpeg, and FluidSynth with a GM soundfont. The software is free and open source.

Spec sheet

What it does
StemForge is a local GPU-accelerated web workstation that separates stems, extracts MIDI, generates audio, composes songs, converts voices, mixes tracks, and exports files.Read 15 Oct 2026 · github.com
Product
StemForge is an open-source, GPU-accelerated audio workstation for stem separation, MIDI extraction, audio generation, and song composition that runs locally in a browser.Read 4 Oct 2026 · github.com
Platforms
The project documents Linux and macOS support, Windows use through WSL, and access through a web browser; Apple Silicon uses MPS, while AMD GPU support via ROCm is untested on real hardware.Read 4 Oct 2026 · github.com
Integrations
The documented workflow connects Separate, MIDI, Synth, Compose, Mix, and Export tabs so outputs can flow between pipelines; MIDI Out connects to hardware synths via the browser's Web MIDI API.Read 4 Oct 2026 · github.com
Privacy
The project says audio stays on the user's computer and no internet connection is required after initial model downloads.Read 4 Oct 2026 · github.com
AMD support
Linux AMD GPUs use a separate ROCm 7.1 project variant, but the project describes end-to-end AMD hardware support as untested.Read 15 Oct 2026 · github.com
Audio generation
Stable Audio Open is used for text-conditioned stereo generation with optional stem, MIDI, and mix conditioning and project support up to 600 seconds through chunking.Read 15 Oct 2026 · github.com
Basic Pitch integration
Spotify describes Basic Pitch as an audio-to-MIDI converter with pitch-bend detection and publishes it under Apache-2.0.Read 15 Oct 2026 · github.com
Documentation
The repository links an INSTRUCTIONS.md guide, requirements, installation steps, launcher flags, model-cache guidance, and a third-party notices file.Read 15 Oct 2026 · github.com
Export formats
Export supports WAV, FLAC, MP3, and OGG, including ZIP downloads.Read 15 Oct 2026 · github.com
Hugging Face gate
The Synth tab requires accepting Stable Audio Open's license and logging in locally with a read token before model access.Read 15 Oct 2026 · github.com
Install
Installation uses a recursive Git clone, uv sync, and uv run python run.py, then serves the interface at localhost port 8765.Read 15 Oct 2026 · github.com
Licence
The repository identifies StemForge itself as Apache License 2.0 and permits commercial use, modification, and distribution.Read 15 Oct 2026 · github.com
Limit: Demucs artifacts
Demucs documentation reports substantial bleeding and artifacts for the piano source in its experimental six-source model.Read 15 Oct 2026 · github.com
Limit: Demucs maintenance
The Demucs repository is archived and says its original maintainer no longer actively maintains it beyond important fixes.Read 15 Oct 2026 · github.com
Limit: licence inconsistency
The v1.0.0 release page instead describes that release as personal and non-commercial under PolyForm Noncommercial 1.0.0, creating a material licence discrepancy.Read 15 Oct 2026 · github.com
Limit: Stable Audio language and style
The model card says performance is weaker outside English, varies across musical styles and cultures, and is generally better for sound effects and field recordings than music.Read 15 Oct 2026 · huggingface.co
Limit: Stable Audio vocals
Stability's model card says Stable Audio Open is not able to generate realistic vocals.Read 15 Oct 2026 · huggingface.co
Limit: third-party rights
The project warns that its Apache licence does not cover third-party model weights and lists some weights as non-commercial, unknown-licence, or requiring separate review.Read 15 Oct 2026 · github.com
Local processing
The project says audio remains on the user's computer and that internet access is unnecessary after model downloads.Read 15 Oct 2026 · github.com
macOS support
Apple Silicon Macs use MPS acceleration, while Intel Macs operate CPU-only.Read 15 Oct 2026 · github.com
Maker identity
The repository owner identifies as Todd (tsondo) and lists Germany as the location.Read 15 Oct 2026 · github.com
MIDI
StemForge extracts MIDI from instrument stems with BasicPitch and vocal melodies with faster-whisper and pitch tracking, and can send MIDI to hardware synths through Web MIDI.Read 4 Oct 2026 · github.com
MIDI browser support
MIDI Out requires Chrome or Edge and a secure browser context; the project documents HTTPS or localhost as supported secure origins.Read 4 Oct 2026 · github.com
MIDI extraction
Instrument MIDI uses Basic Pitch, while vocal MIDI uses faster-whisper with pitch tracking; FluidSynth provides browser previews.Read 15 Oct 2026 · github.com
MIDI Out
Extracted MIDI can be sent to software or hardware synthesizers through Web MIDI, with program changes and a panic control; Chrome or Edge is required by the project.Read 15 Oct 2026 · github.com
Mixing
Mix combines audio stems, MIDI renders, synth outputs, and composed songs with per-track instruments and volume, then renders FLAC.Read 15 Oct 2026 · github.com
Mixing and export
The Mix tab combines audio stems, MIDI-rendered tracks, synth outputs, and composed songs; exports support WAV, FLAC, MP3, and OGG, individually or as a ZIP archive.Read 4 Oct 2026 · github.com
Model download burden
First use can download about 300 MB per Demucs model, 500 MB per BS-Roformer model, 2 GB for Stable Audio, and about 20 GB for AceStep.Read 15 Oct 2026 · github.com
Model downloads
Models download lazily on first use; the Compose tab's AceStep models require about 20 GB, and the Stable Audio Open model requires HuggingFace license acceptance and authentication.Read 4 Oct 2026 · github.com
Notable issue
An open July 20, 2026 issue asked for AMD ROCm support and Python 3.12 compatibility.Read 15 Oct 2026 · github.com
Notable release
The repository lists v1.0.0 as its latest release, published February 24, with a hybrid generative workstation description.Read 15 Oct 2026 · github.com
Repository activity
The GitHub repository shows 37 stars, 6 forks, two open issues, and three open pull requests when opened.Read 15 Oct 2026 · github.com
Requirements and limits
The documented setup requires Python 3.12, FFmpeg, and FluidSynth with a GM soundfont; NVIDIA GPU acceleration requires driver 580 or newer, while CPU-only operation is significantly slower.Read 4 Oct 2026 · github.com
RVC dependency
StemForge vendors Applio for RVC inference; Applio identifies itself as MIT-licensed and says its future development is mainly security and dependency maintenance.Read 15 Oct 2026 · github.com
Security and licensing
StemForge is licensed under Apache 2.0, while the project notes separate upstream model license terms, including Stability AI's Community License for Stable Audio Open and no specified license for one six-stem BS-Roformer model.Read 4 Oct 2026 · github.com
SFX builder
The Synth tab includes a timeline canvas for arranging clips with fades, volume, reference alignment, and rendering into Mix.Read 15 Oct 2026 · github.com
Song composition
AceStep supports create, rework, Lego, and complete modes plus LoRA and LoKR adapter training.Read 15 Oct 2026 · github.com
Song composition and voice conversion
The Compose tab uses AceStep for song generation and RVC for voice conversion, with 14 built-in voices and a HuggingFace model browser.Read 4 Oct 2026 · github.com
Stem separation
Demucs and BS-Roformer can split recordings into vocals, drums, bass, other, guitar, and piano stems.Read 15 Oct 2026 · github.com
System requirements
The documented stack requires Python 3.12 through uv, FFmpeg 5.1 or newer with development headers, FluidSynth with a GM soundfont, and recommends an NVIDIA driver version 580 or newer for CUDA 13.Read 15 Oct 2026 · github.com
Third-party model licensing
Stable Audio Open 1.0 is marked stable-audio-community on Hugging Face and requires separate commercial licensing consideration.Read 15 Oct 2026 · huggingface.co
Vocal enhancement
Enhance provides eight UVR cleanup presets and a CREPE plus Praat PSOLA tuning mode; an effects mode is marked planned.Read 15 Oct 2026 · github.com
Voice conversion
RVC voice conversion offers fourteen built-in voices, Hugging Face model search, pitch shifting, and selectable F0 methods.Read 15 Oct 2026 · github.com
Web MIDI limitation
MDN classifies Web MIDI as limited-availability, requiring a secure HTTPS context and user permission.Read 15 Oct 2026 · developer.mozilla.org

Plans and prices

Local StemForge Free self-hosted · no subscription · no per-track fees · model downloads required Read 15 Oct 2026 · github.com
StemForge Free No subscriptions or per-track fees Open-source; runs locally on your hardware Read 4 Oct 2026 · github.com

Every StemForge plan and price →

Strong points and limits

Strong points

  • Separates six listed stem types.
  • Exports WAV, FLAC, MP3, and OGG.
  • Audio stays on-device, according to the project.
  • Free, open-source Apache 2.0 software.

Limits

  • Setup requires Python 3.12, FFmpeg, and FluidSynth.
  • CPU-only operation is significantly slower.
  • AceStep models require about 20 GB.
  • MIDI Out requires Chrome or Edge in a secure context.

Who it suits

StemForge suits musicians and audio makers who want connected separation, MIDI, generation, composition, and mixing workflows running locally. It requires Python 3.12, FFmpeg, and FluidSynth; CPU-only operation is significantly slower.

Verdict

StemForge brings several audio workflows together while keeping audio local after initial model downloads. Check its setup requirements, model downloads, and separate upstream model license terms before using it.

Where Specifiction places it

Alternatives to StemForgeAll alternatives →