Looking for the product overview? See the Media Asset Management service page and book a walkthrough from there.
Every asset has a spine. Give yours a signal.
Spine & Signal is a Media Asset Management (MAM) platform for anyone with a growing library of video, audio and stills to organise — from decades-deep broadcast archives to the fast-moving footage pipeline behind a YouTube channel or studio. Search, structure, encode and govern it all from one modern system.
Why teams switch
A media library is built to outlive the software it is catalogued in. The software rarely outlives the browser plugins it was built on:
- Dead on arrival — embedded players and editors built on ActiveX, Flash or Silverlight simply don't load in a current browser; features that worked for a decade go dark overnight.
- Unenforced access — "permission tags" that only hide a button are not access control. If the action isn't re-checked server-side, anyone who can reach the URL can perform it.
- Fragile credentials — plaintext passwords in a decades-old user table are a liability every day they are live.
Spine & Signal was designed as the modern replacement for exactly this kind of legacy Java MAM: a Blazor web application, a governed REST API layer, queue-driven workflow workers, and a fleet of media-intelligence services that understand audio, video and stills the way an archivist — or an editor — does.
One library, six disciplines
01 · Discovery — search built on real structure Per-type predicate search across masters, versions, stems, stills and generic audio/video; full-text phrase search across synced metadata; saved, reopenable queries you can hand to the team.
02 · Asset intelligence — every asset, fully instrumented Interactive waveform playback direct from source media; edit-decision-list and segment editing across physical and logical versions; full structure and relations at a glance.
03 · Audio pipeline — archive-grade audio, verified FLAC/MP3/AAC encoding with bit-exact 24-bit preservation and built-in integrity verification; large-format (>4 GB) WAV handling with corrupt-header repair; automatic track-boundary and scene segmentation, speech/music classification, spoken-language identification.
04 · Video & image — vision, not just storage Automatic proxy generation and scene-detected keyframes; timecoded speech-to-text matched to each detected shot; broadcast QC in a single pass (black/frozen/silent frames plus EBU R128 loudness); OCR with language detection; face detection, near-duplicate detection via perceptual hash and AI semantic similarity; no-reference image quality scoring.
05 · Operations — run it like an instrument panel Live queue-message admin with filtering, resubmission and failure tracking per process; workflow visibility with active-instance counts; endpoint topology and system health monitoring.
06 · Governance — access that's actually enforced Single sign-on via your existing directory (Windows AD / Kerberos); role-gated administrative surfaces; a structured, queryable audit trail on every automated processing step.
Editorial AI — Claude proposes, your team decides
The analysis services produce the facts about every asset: timecoded transcripts, OCR text, scene keyframes, classifications and QC flags. On top of that, Spine & Signal adds an editorial layer powered by Claude (Anthropic). It turns those facts into the work an archivist or editor would otherwise do by hand.
- Tag suggestions from your own vocabulary — Claude proposes keywords for each asset, restricted to your controlled vocabulary, each with a confidence score and the evidence it relied on. Free-text tags outside the vocabulary are rejected by design. Available now.
- Titles, teasers and descriptions — draft editorial fields generated from transcript, scenes and existing metadata. In development.
- Translation — metadata translated between German, English, French and Spanish, with the source language detected from the audio. In development.
- Captions — subtitles built from the timecoded transcript, tidied and translated, with the original timing kept. In development.
- Reel drafts — a proposed cut list of scenes with in/out points and captions, which becomes a real edit-decision list once accepted. In development.
- Natural-language search — describe what you are looking for in plain words; Claude turns it into a structured query you can see and adjust, combined with visual similarity search over the keyframes. In development.
A person signs off on everything. Nothing an AI model produces goes live on its own. Every result is stored as a proposal next to the asset, with its evidence. An editor accepts or rejects it in the asset view, and only accepted proposals become real metadata, each with an audit entry.
Your data, your choice of model. Claude only receives a compact text summary of the asset: metadata, transcript and analysis results. It never receives the media files themselves. The model sits behind a single interface, and Spine & Signal supports two today: Claude via the Anthropic API, and local models via Ollama running on your own hardware. Same tasks, same workflow, same review step. With Ollama, no asset data leaves your network.
Platform figures
- 9 structured search catalogues across footage, masters, stills and audio
- 3 media types natively analysed — audio, video, still image
- 3 archive-grade audio codecs, encoded and verified — FLAC, MP3, AAC
- 24-bit source fidelity preserved bit-exact through the encoding pipeline
Under the hood
The platform is built entirely on .NET: a Blazor Web App frontend, a REST API over the archive database (Oracle or SQL Server), queue workers and a workflow engine for automated processing, an encoding service for archive-quality audio, and three media analysis toolchains — for audio (ManagedBass), video (FFmpeg) and stills (ImageMagick/Tesseract) — extended with Whisper speech recognition and CLIP vision models, and an editorial AI layer that runs on Claude or on local models via Ollama. Every analysis result is written under one persistence contract, so audio, video and image findings land in the same queryable store.
Two ways to run it
Spine & Signal Cloud (SaaS) — we host, patch and scale the platform. Managed upgrades, elastic encoding and analysis capacity at peak ingest, security and SSO onboarding from day one. The fastest path from decision to first indexed asset.
Spine & Signal On-Prem (Installation) — we install and configure the platform inside your infrastructure, alongside your existing storage and directory services. Full data residency, fits air-gapped and regulated networks, operated by your own team after a guided handover.
Same platform, same feature set — the difference is who operates it. Get in touch to book a walkthrough.