Quality assurance · Production

Ollang scales media localization with 76% less manual processing after integrating AssemblyAI Voice AI

The problem

Ollang's multi-agent localization platform required exceptionally accurate transcription as its foundational first step, but existing cloud providers delivered insufficient accuracy for non-English audio—with poor punctuation and capitalization handling that cascaded errors through every downstream workflow and threatened Ollang's value proposition with streaming and broadcast clients.

First attempt

Prior cloud transcription providers delivered insufficient accuracy for non-English audio, with poor punctuation and capitalization handling that cascaded errors throughout Ollang's entire multi-agent workflow.

Workflow diagram · grounded in source
1
Client video submitted
trigger
“Media clients working with extensive video libraries—long interviews, industry events, and hundreds of series episodes—demanded production-ready results without compromise”
2
Universal Speech-to-Text transcription
ai_action
“The state-of-the-art Universal Speech-to-Text model boasts more than 93.3% accuracy, even on noisy audio, providing the industry's lowest word error rate (WER). The model also supports multilingual transcription across numerous languages…”
3
Multi-agent orchestration
ai_action
“agentic AI that dynamically selects the best models and continuously self-corrects for enhanced performance across languages”
4
Localization output delivery
output
“from captioning and subtitle translation to dubbing workflows—enabling the system to make faster, more confident automated decisions”
5
Optional human review
human_review
“More clients now place AI dubbing orders without requesting human review, reflecting the improved reliability of Ollang's end-to-end platform”
Reported outcome

After integrating AssemblyAI's Universal Speech-to-Text API, Ollang achieved a 76% reduction in human-in-the-loop effort, a 30-40% improvement in overall platform accuracy, 97%+ production-ready results for most content types, and a 25% increase in autonomous service orders.

Reported metrics
Reduction in human-in-the-loop effort76%
Improvement in overall platform accuracy30-40%
Production-ready results without human intervention97%+
Increase in autonomous service orders25%
Show all 5 reported metrics
reduction in human-in-the-loop effort76%
improvement in overall platform accuracy30-40%
production-ready results without human intervention97%+
increase in autonomous service orders25%
AssemblyAI model transcription accuracymore than 93.3%
Reported stack
AssemblyAIUniversal Speech-to-Text API
Source
https://www.assemblyai.com/customers/ollang-customer-story
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

After integrating AssemblyAI's Universal Speech-to-Text API, Ollang achieved a 76% reduction in human-in-the-loop effort, a 30-40% improvement in overall platform accuracy, 97%+ production-ready results for most conte…

What tools did this team use?

AssemblyAI, Universal Speech-to-Text API.

What results were reported?

Reduction in human-in-the-loop effort: 76%; Improvement in overall platform accuracy: 30-40%; Production-ready results without human intervention: 97%+; Increase in autonomous service orders: 25% (source-reported, not independently verified).

What failed first in this deployment?

Prior cloud transcription providers delivered insufficient accuracy for non-English audio, with poor punctuation and capitalization handling that cascaded errors throughout Ollang's entire multi-agent workflow.

How is this quality assurance AI workflow structured?

Client video submitted → Universal Speech-to-Text transcription → Multi-agent orchestration → Localization output delivery → Optional human review.

Related quality assurance cases