SpectraCleanse AI

SpectraCleanse AI strips AI provenance markers and injects SEO metadata to beat algorithmic suppression.

Upload your audio or video file, choose a platform preset, and SpectraCleanse AI will forensically wipe every embedded tag that signals AI-generated origin — then inject clean, platform-optimised metadata powered by Gemini to maximise your reach on YouTube, Spotify, Apple Music, TikTok, and beyond.

---

Try it free

→ spectracleanse.com

No credit card required. Free accounts include 3 files per month. Upgrade to Creator ($9.99/mo) or Studio ($24.99/mo) for unlimited processing and batch uploads.

---

How it works

1. Upload — drag in any MP3, WAV, FLAC, M4A, or MP4 file (up to 500 MB).
2. Analyse — the forensic engine reads every embedded tag and identifies provenance markers.
3. Cleanse — a nuclear wipe removes all XMP, IPTC, and ID3 fields that could trigger algorithmic detection.
4. Inject — Gemini generates an SEO-optimised title, description, and tag set tuned to your chosen platform.
5. Download — receive a clean file with a full forensic report showing exactly what was removed.

---

Open source

The source code is available at github.com/ChrisAdamsdevelopment/SpectraCleanseAI and is released under the MIT License. You are free to self-host, fork, or contribute.

---

Local development notes

• Backend defaults to developer-friendly mode when NODE_ENV is not production.
• If Stripe env vars are missing locally, /api/create-checkout-session can return a mock checkout redirect (set ENABLE_MOCK_CHECKOUT=true).
• In production, Stripe variables are still required and the server will fail fast if they are missing.

---

QA / Deployment Testing

For a step-by-step manual validation flow (local, API smoke, auth, billing, upload, cleanse, Docker, and production readiness), see docs/manual-qa-checklist.md.

• Browser metadata analysis uses maintained music-metadata with graceful fallback (parseError) when parsing fails, times out, or is skipped for very large files.
• Quick Cleanse metadata writing remains local/browser-side (MP3 via browser-id3-writer).
• Full Server Cleanse runs through /api/process; MP3 requests are rejected with HTTP 422 and guidance that Full Server Cleanse is best supported for MP4/M4A, while WAV/FLAC may still be rejected when ExifTool cannot safely rewrite them.

---

Contact

Questions, partnerships, or enterprise enquiries: hello@spectracleanse.com

---

Native Node deployment runtime

• Native Render/Node deployments should use Node 20.20.2 (recommended) or a Node version within the supported engines range: >=18 <23.
• If Render defaults your service to a newer Node release, set NODE_VERSION=20.20.2 in the service environment.
• Node 24 is currently not supported for native installs in validation because better-sqlite3 native compilation failed under Node 24.
• Docker deployments already pin Node 18 via the repo Dockerfile.

---

Docker production deployment

This repository includes a multi-stage Dockerfile that builds the frontend and packages dist/ into the final runtime image so server.js can serve the SPA in production.

Build image

docker build -t spectracleanseai:latest .

Run container

docker run --rm -p 3001:3001 \
  -e NODE_ENV=production \
  -e JWT_SECRET=your_jwt_secret \
  -e STRIPE_SECRET_KEY=sk_live_xxx \
  -e STRIPE_WEBHOOK_SECRET=whsec_xxx \
  -e STRIPE_CREATOR_PRICE_ID=price_xxx \
  -e STRIPE_STUDIO_PRICE_ID=price_xxx \
  -e GEMINI_API_KEY=your_gemini_api_key \
  -e FRONTEND_URL=https://your-frontend-domain.example \
  -e DB_PATH=/data/spectra.db \
  -v spectracleanse_data:/data \
  spectracleanseai:latest

Required production environment variables

• NODE_ENV=production
• JWT_SECRET
• STRIPE_SECRET_KEY
• STRIPE_WEBHOOK_SECRET
• STRIPE_CREATOR_PRICE_ID
• STRIPE_STUDIO_PRICE_ID
• GEMINI_API_KEY
• FRONTEND_URL
• DB_PATH
• REDIS_URL (only if your deployment still uses Redis externally)

Stripe vs local mock checkout

• Local development may use ENABLE_MOCK_CHECKOUT=true when Stripe variables are not set.
• Production must use real Stripe configuration; do not rely on mock checkout in production.

Never commit real secrets to source control.

Batch processing API

• POST /api/process-batch (authenticated): processes up to 20 uploaded files sequentially for paid plans (Creator/Studio). Free plan returns 403.
• GET /api/download/:token (authenticated): one-time secure download for batch outputs.
• MP3 server cleanse remains unsupported (422 for single process, per-file error in batch); use Quick Cleanse Browser for MP3. Full Server Cleanse is best supported for MP4/M4A, and WAV/FLAC may be rejected if ExifTool cannot safely rewrite them.
• Batch requests enforce a 2GB post-upload soft guard; production deployments should still enforce proxy/body-size/disk limits.