VoiceRule Launches Polita, AI- and Expert-Powered Dubbing Service

VoiceRule, an AI-based content localization and creator support solutions company, announced on the 17th the launch of Polita(Polita), a B2B dubbing production service that combines AI and experts. Polita is a finished-delivery service in which AI generates translations and dubbing, after which experts proofread and review the entire output. Through Polita, VoiceRule aims to leverage AI’s production speed and cost efficiency while improving translation quality, pronunciation quality, emotional expression quality, and sync quality through expert review.

VoiceRule introduced Polita as a service that preserves the advantages of creating dubbing with AI alone while having experts refine the entire output to improve quality. Polita is designed to use AI’s fast production speed and cost efficiency while raising dubbing quality through expert review.

In particular, during AI dubbing for videos with multiple speakers, speaker recognition errors can occur, and in such cases, lines may be generated in another character’s voice. Accordingly, Polita applies speaker identification through video analysis and directly inputs speaker information into the AI engine to help prevent speaker recognition errors at the generation stage.

Depending on the nature of the project, different voice generation and conversion methods, as well as voice assets, are applied. Project-specific elements include zero-shot STS(Speech-to-Speech) and a proprietary voice model library. The zero-shot STS feature does not require separate training and generates desired text speech based on the original audio.

For content where the original speaker’s performance and emotional expression are important, a voice conversion method that reflects breathing and emotional nuance is applied. For content where information delivery is more important, such as lectures and e-learning, a method is used that assigns an appropriate voice to each speaker.

VoiceRule has localized a cumulative 300,000 pieces of content and uses accumulated translation data and an infrastructure of more than 1,300 language workers for dubbing production. VoiceRule emphasized that language experts can adjust dubbing scripts, and in the process they take into account translation accuracy, lip movement, and utterance length.

After that, the company selects the best result from among multiple AI-generated dubbing voices and automatically adjusts volume and sync. Review can proceed through up to 5 rounds, and a PM with a background as a sound engineer performs the final check on sync, sound quality, and tone. Only the output that passes final review is delivered.

Polita supports a total of 70 languages, including English, Japanese, Chinese, Spanish, French, Vietnamese, Thai, and Indonesian, and the support method for all 70 languages follows the same production process. Costs are reduced by 60% to 80% compared with professional voice actor recording, and production time is shortened to about one-fifth compared with studio production. Videos of around 30 minutes can be produced in about 5 business days.

VoiceRule CEO Lee Sang-hun said that, as overseas distribution expands, the importance of rapidly producing multilingual content is becoming more pronounced. He explained that, beyond speed, preserving the original work’s meaning and emotion, as well as maintaining consistency in character voices, is also important. He added that Polita plans to support the global expansion of diverse content by combining AI technology with VoiceRule’s localization and sound production capabilities.

Source: TECHWORLD · Lee Kwang-jae
Original: https://www.epnc.co.kr/news/articleView.html?idxno=407047

References

• https://www.epnc.co.kr/news/articleView.html?idxno=407047