# Competitive reference offerings and differentiation

**No bidder or incumbent has been confirmed for this opportunity.** The offerings below are public references that could inform a competitor, a component choice, a baseline or a partner. Vendor descriptions establish documented capabilities, not independent quality results, contract suitability or intent to bid. An undocumented feature is unknown here, not proven absent.

## What established offerings already do

| Public reference | Documented capability | What we need to prove in our response |
|---|---|---|
| [Hindenburg Narrator tutorials](https://hindenburg.com/academy/lessons-abc/narrator-tutorials/) and [feature comparison](https://hindenburg.com/products/audio-books/comparison/) | Explicit NLS DTB export, alongside other accessible formats and navigation/markup tools | Do not claim others cannot make NLS books. Demonstrate the exact approved profile, protected output, AWS workflow, delivery rights and measured operator effort for our proposed system. |
| [DAISY Pipeline](https://daisy.github.io/pipeline/Get-Help/User-Guide/Text-To-Speech/) | TTS-assisted conversion, synchronized audio/text and pronunciation lexicons, using local and cloud engines | Explain reuse versus custom work and verify NLS-specific protection/packaging. Local inference and lexicons alone are not novel. |
| [DAISY Obi](https://daisy.org/activities/software/obi/) | Open-source structured audio and accessible-publication production, editing, conversion and validation | Benchmark reviewed commercial-audio navigation and operator correction. Source availability alone does not distinguish us. |
| [Pratsam Producer](https://pratsam.fi/pratsam-producer-product.html) | Automated DAISY/EPUB speech-synthesis production, content comparison, manual review and cloud or dedicated service | Existing market offerings already combine automation and QA. Demonstrate our own long-book fidelity, correction economics and Library handover terms. |
| [Amazon Polly](https://aws.amazon.com/polly/features/) as an integrator's component | Long-form voices, lexicons and speech marks; [asynchronous synthesis](https://docs.aws.amazon.com/en_en/polly/latest/dg/asynchronous.html) writes results to S3 | A voice service is not a complete DTB system. Compare accepted-output quality and lifecycle dependencies; do not presume open models are cheaper or that paid APIs meet the draft's rights terms. |

Sources checked October 2, 2026. None establishes another provider's proposed price, staff, exact protected-NLS compatibility, contract rights or approved deployment for this acquisition. Confirm licensing and test access before evaluating any product.

## Compete through inspectable evidence

Propose a common corpus and acceptance tasks that apply fairly to any supplier: produce protected samples from both routes; navigate on actual target players; account for source content; repair a seeded defect; measure review and compute effort; recover interrupted work; and install the delivered system independently. These tests turn claims into comparable results while leaving implementation choices open.

| Comparison dimension | What to record | Avoid this misleading shortcut |
|---|---|---|
| Audio fidelity and listener suitability | Whole-book defects, fatigue/clarity feedback, voice stability and final decoded audio | A single polished short clip |
| Navigation and protection | Same profile, metadata expectations, player matrix and actual protected artifacts | A generic DAISY label or browser playback |
| Correction effort | Reviewer minutes, repair time, fraction regenerated and missed defects on the same seeded issues | An unsupported percentage saving |
| Cost | All labor, tool, compute, review, failure/rework, support and warranty assumptions | GPU cost compared with a vendor's complete service price |
| Independence and rights | Actual continuing-use/redistribution terms and a clean handover test | “Open weights” treated as ownership of all components |
| Integration and reliability | Approved interface transactions, retries, restore and audit trail | AWS hosting treated as complete integration or security authorization |

Do not publish a scored vendor league table until comparable tests exist. Mark untested as untested and preserve failed runs. Future evaluation weights must come from the solicitation, not a favorable scorecard we invent.

## Reuse and teaming decision

Before building each subsystem, evaluate whether an established component can satisfy the profile, runtime, rights and maintainability requirements. Reuse can reduce implementation work; integration and regression testing still cost effort. A qualified partner may close protected-DTB, accessible-reader testing or delivery-record gaps faster than another model experiment. No vendor named above is a committed partner.

Our proposed differentiator is the demonstrated contract-specific delivery chain and its correction/handover evidence. The [strategy page](win-strategy.md) connects that position to buyer outcomes; the [gap register](gaps.md) records why it is not yet a claim of superiority.