Lyvora Sounds
Audio Processing Pipeline
Reliable audio processing, from ingestion to delivery.
Selket engineered an audio-processing pipeline that coordinates media processing, voice-tag handling, copyright verification and asset delivery across a connected music platform.
The work combines backend services, automated processing, storage infrastructure and platform reliability to make audio publishing more dependable.
The Challenge
Audio processing depends on several components working together. When processing configuration conflicts with validation rules, a single failure can prevent previews, voice tags and copyright checks from completing.
The engineering challenge was to make these components work reliably while improving performance, controlling infrastructure costs and maintaining a platform that could evolve without a disruptive migration.
How the Pipeline Works
- 01Audio ingestion
- 02Transcoding & voice tags
- 03Preview & copyright check
- 04Storage & delivery
01. Audio Ingestion
Audio enters the platform through its processing workflow, where backend services coordinate the work required to prepare assets for publishing.
02. Transcoding and Voice Tags
The processing service handles audio transcoding and voice-tag generation. Configuration must remain within the validation rules expected by the processing service.
03. Preview Generation and Copyright Verification
The workflow supports preview generation and AudD copyright verification. An issue with the voice-tag volume configuration had caused processing requests to fail; correcting the configuration restored successful processing and these downstream functions.
04. Storage and Delivery
Audio files, stems and licences are served through Cloudflare R2, using presigned URLs. This reduces reliance on the previous storage delivery path and eliminates the reported cloud egress fees on audio downloads.
The Engineering Solutions
Processing Reliability
The voice-tag volume was constrained to the supported range, resolving HTTP 422 processing failures and restoring successful audio processing.
- Voice-tag volume outside the supported range
- HTTP 422 processing failure
- Previews, voice tags and copyright checks blocked
- Volume constrained to the supported range
- Processing succeeds
- Previews and copyright checks restored
Connected Backend Architecture
A Nuxt Strangler Proxy routes selected requests to a centralised backend API while allowing unmigrated routes to continue using the existing implementation. This enables incremental migration rather than an all-at-once replacement.
Faster Database Connectivity
Cloud Run services were relocated to Europe to sit closer to the Neon database region. The report records a reduction in connection handshake time from approximately 400 ms to 10 ms.
About 40× faster to establish a database connection.
More Efficient Asset Delivery
The platform migrated audio, stems and licences to Cloudflare R2, with compatibility for the previous storage service during the transition.
- Audio, stems and licences
- Previous storage delivery path
- Cloud egress fees on downloads
- Audio, stems and licences
- Cloudflare R2 via presigned URLs
- Reported egress fees eliminated
Security and Automated Testing
The engineering work included protecting deployment secrets, removing a client-side admin bypass from production builds and adding a Playwright end-to-end test suite covering key user journeys.
The Outcome
The September 2026 engineering report records improvements across audio-processing reliability, database connectivity, asset delivery, migration flexibility and deployment security.
- Processing reliabilityHTTP 422 failures resolved; audio processing restored.
- Database connectivityHandshake time from ≈ 400 ms to 10 ms.
- Asset deliveryAudio, stems and licences served from Cloudflare R2.
- Migration flexibilityRoutes migrate incrementally through a strangler proxy.
- Deployment securitySecrets protected, admin bypass removed, end-to-end tests added.
Together, these changes strengthened the infrastructure supporting the audio platform—not just the individual processing step, but the services required to deliver the resulting assets reliably.
Built for Connected Systems
Audio processing is only one part of a dependable platform. Processing services, databases, storage, APIs and security controls must work together as a coherent system.
Selket approaches these challenges through practical engineering: identifying failure points, improving the connections between services and building infrastructure that supports reliable operations.
Building a system with similar challenges?
Discuss your requirements with Selket to identify the right integration, automation and infrastructure approach.