Audio and podcasting decision
Moises
A competent developer can reproduce a usable vocal-removal web tool (upload → separation → mixer → download) using open-source separation models in about a week and modest monthly GPU costs, but reproducing Moises’ full multi-platform polish, proprietary models, licensed-data advantages, and large user ecosystem is not realistic for a lone builder.
Visit website↗No open-source build does this yet
Nothing published replaces this one, so a replacement starts from an empty file. Here is what it would have to cover.
What a replacement has to do
- Upload audio/video → run stem-separation model → present isolated stems in a mixer → allow mute/download/export of stems
What it still won’t have
- Proprietary trained models and any licensed training data
- Polished multi-platform apps (iOS, Android, Desktop) and extensive UX polish
- Large user community and brand recognition
- Any paid-plan features like Hi‑Fi separation limits or integrated cloud storage
What remains hard
- Brand trust
Trusted by +75 million artists worldwide
- Brand trust
Recognized by Apple as the iPad App of the Year
First-year cost
Keep paying
Paying is—cheaper in year one.
On cash alone, building overtakes the subscription at 52 seats.
Money you would actually spend
Time you would spend
—
What you would spend
What we assumed
The verdict above measures whether you could build it. This one is only about money.
Runnable build prompt
Build a minimal web service in Node.js + Next.js with a PostgreSQL DB and Redis queue that lets a user upload an audio/video file, runs open-source source-separation (e.g., Demucs or a provided container) on a GPU worker, stores resulting stems, provides a frontend mixer to play/mute/solo stems, offers pitch and speed controls, chord/BPM detection, and a download/export button for MP3/WAV. In scope: upload API, background worker orchestration (Docker + GPU on a cloud VM), basic authenticated UI, file storage on S3, notifications for job completion. Out of scope: native mobile apps, paid billing integration, large-scale user management, training new models. Include error handling, retries for worker failures, automated tests for upload + job completion, and a deployment script + README to run on a single cloud VM with GPU.
How we checked
How the score was reached
- Partly verdict base52
- 3 cited sources+3
- Evidence score55
The base comes from the verdict. Everything under it is a check that either happened or did not, and each one is a fact frozen in this record rather than a judgement made at render time — so the same evidence always produces the same number.
How scoring works →Cited sources · 3
Every page the run actually retrieved.
- official productMoises home
- official docsMoises features — Vocal Remover
- official pricingMoises pricing / home
Integrity checks
What held up, and what did not.
