From Captions to Translations: Evaluating Readiness for AI-Powered Live Language Solutions

Metadata

Publisher: SMPTE — Pasadena, CA
Doc Type: Conference Paper
Content Type: Original Research
Volume: 00, pp. 1–18
Abstract: As artificial intelligence reshapes live media workflows, the focus of new projects in language and localization is entering a new phase - shifting from same-language AI captioning to new applications for real-time language translation. This paper discusses the merits of the Number Translation Recognition (NTR) metric for evaluating translations, and investigates through this metric whether the performance trajectory of low latency automated translation is on track to follow the same path of rapid improvement seen in the similar Number Edition Recognition (NER) scores commonly used to measure the accuracy of same language captioning. We begin with a comparative analysis of NER and NTR metrics at similar stages of model maturity. Drawing on historical NER data - where captioning accuracy has steadily improved from 97 - 97.5% in 2021 to 98.5 - 99% in 2025 - we present early NTR evaluation data and identify parallel performance patterns. We argue that, like NER, NTR can benefit from structured evaluation frameworks that enable clear benchmarking and drive iterative model development. Building on this foundation, the paper explores the operational implications of growth in AI captioning quality, including advances in model evaluation and analytics workflows. Through the use of AI-Media's LEXI (AI captioning) and LEXI Voice (AI translation) platforms as applied use cases, we analyze how structured evaluation datasets and feedback loops enable rapid iteration and measurable quality improvements across AI-based language products. Finally, the paper proposes a framework for evaluating industry readiness to adopt AI-powered live translation as a production-standard solution. This framework considers technical performance benchmarks, workflow integration, and operational preparedness - providing practical guidance for broadcasters, content owners, and live event producers seeking to expand reach through multilingual delivery.
Publication Date: 2025-10-13
ISBN: Electronic: 978-1-61482-966-9
Author(s): Bill McLaughlin
Cody Farenga
Cody Parnell
Keyword(s): AI Captioning, AI Translation, NER, NTR, Live Automatic Captions, Live Translation, Automatic Speech Recognition, Machine Translation, Text-to-Speech, Large Language Models (LLMs), Voice Synthesizer, Evaluation Metrics, Structured Datasets, Broadcast Accessibility, Multilingual Delivery, Real-Time Media Workflows, Quality Benchmarking, Iterative Model Development, Workflow Integration, Operational Readiness, Compounding AI Workflows
Copyright: © 2025 SMPTE

Bibliographic Reference(s)

Source Data (JSON)

Full registry record with provenance metadata. Open directly: /api/doc/978-1-61482-966-9-8.json

Reference Tree

Explore all references and references to this document, as a navigable tree.

Open Reference Tree

Reference this Doc

Plain text (ISO 690 compliant)

Preview:

Bill McLaughlin, Cody Farenga, and Cody Parnell; From Captions to Translations: Evaluating Readiness for AI-Powered Live Language Solutions, MTS 2025, Article 8 (pp. 1 to 18); SMPTE, 2025, ISBN: 978-1-61482-966-9

Snippet:

Bill McLaughlin, Cody Farenga, and Cody Parnell; From Captions to Translations: Evaluating Readiness for AI-Powered Live Language Solutions, MTS 2025, Article 8 (pp. 1 to 18); SMPTE, 2025, ISBN: 978-1-61482-966-9

HTML (ISO 690 compliant)

Preview:

Snippet:

<span class="citation">Bill McLaughlin, Cody Farenga, and Cody Parnell; <cite>From Captions to Translations: Evaluating Readiness for AI-Powered Live Language Solutions</cite>, MTS 2025, Article 8 (pp. 1 to 18); SMPTE, 2025, ISBN: 978-1-61482-966-9</span>

SMPTE's HTML Pub

Preview:

Snippet:

<li>
Bill McLaughlin, Cody Farenga, and Cody Parnell; <cite id="bib-978-1-61482-966-9-8">From Captions to Translations: Evaluating Readiness for AI-Powered Live Language Solutions</cite>, MTS 2025, Article 8 (pp. 1 to 18); SMPTE, 2025, ISBN: 978-1-61482-966-9
</li>