The use of AI technology and machine learning continues to grow so does the amount of connected devices consumers and businesses can choose from. The core technology behind these connected devices is speech and voice recognition. It is estimated that the global voice recognition market will hit $30 billion by 2026. Natural Language Processing (NLP) is at the heart of voice recognition technology. Voice recognition technology is used in digital assistants such as Alexa and Siri, which can be run on various devices, from smartphones to speakers. In addition, it is vital in speech-to-text applications. Voice AI startups in Korea are developing innovative tools that are not only disrupting the industry in Korea. As more and more households in Korea use voice assistants and voice commands, it will become a part of their daily lives.
Therefore, with increased adoption and use, voice AI startups in Korea will look to capitalize on this up-and-coming market globally. While consumers are a great target, there is a rise in B2B software and APIs that offer distinct competitive advantages and generate fast ROI for various industries.
Top Voice AI Startups in Korea
42Maru
Korean AI Startups – 42 Maru is the creators of a QA (Question Answering) system based on AI (Deep Learning) that understands a question and delivers a single correct answer from a large unstructured data set in less than a second. 42Maru can find an answer in real-time using MRC (Machine Reading Comprehension), and as a public reference, they have co-ranked #1 in Standford University’s MRC competition (SquAD 2.0) with Google AI team (as of Nov. 2018).
Their patented QA system is the only deep-learning-based commercial version in Korea. It is already used in over 50% of the smart speakers in Korea. Furthermore, they are working with major Korean companies like LG U+, Daewoo Shipbuilding, KIA Motors, and KT, just to name a few. The platform accurately understands the intention of a user’s question and gives a correct answer based on their data. So far, they have raised $1.7 million in seed funding from KDB Bank.
Their focus moving forward will be to improve their deep learning system and partner with companies dealing with smart speakers, chatbots, connected cars, and data warehouses.
“42Maru is preparing for expansion into European markets, bringing its technology to life for new and exciting customer experiences,” said the CEO of 42 Maru, D.K. Donghwan Kim.
LOVO.AI
LOVO.AI is a voice AI startup in Korea that makes AI-generated voices using AI technology. They work with companies looking to incorporate voice AI into their marketing campaigns or customer service. Their voice AI can read out any script, which is great for call centers, educational centers, and content-producing companies. The company’s focus for 2022 will be to apply its voice AI to the NFT industry. Their NFT for voiceovers for the popular game OVERWATCH sold out on the OpenSea marketplace. This showed that voice actors could generate additional sources of income using NFTs. LOVO.AI will collaborate with other games and metaverse platforms.
XL8.AI
As with the rise in popularity of Korean films and dramas, there is a growing demand for proper subtitles. XL8.AI is an AI startup in Korea specializing in media translation using AI technology. They aim to be the go-to player in the media content industry, from dramas to films and even the gaming industry. Their AI technology can translate content at a much faster rate than humans. It can currently translate over 10 million characters daily and, more importantly, with high accuracy. Moreover, it can be integrated with global media services like Netflix, Disney+, HBO, Hulu, etc. They aim to remove the language barrier by using their cloud solution, which is the same as their company name, XL8.AI.
Baikalai
Baikalai is an AI startup with its proprietary Korean Natural Language Processing engine. Their core three products are broken down below.
deeq BRAIN
deeq Brain is a deep learning technology to discover dementia in the human brain and cognition activity. In particular, this technology analyzes spontaneous speech, representing a slight symptom of cognitive impairment before manifesting degenerative diseases of the nervous system such as Alzheimer’s disease, Parkinson’s disease, or vascular dementia in the earlier stage of MCI(mild cognitive impairment).
deeq NLP
deeq NLP is a Korean natural language process and morphological analysis SW engine developed by Baikal AI. Most data on the Internet is still written in text. The core skill for processing text data is a morphological analysis of natural language processing. Based on their proprietary technology, they have provided NER(named entity recognition) and a hierarchical question-answering system in addition to pre-trained models of BERT and ELECTRA.
deeq STUDIO
deeq Studio is one of the best tools which fulfill the requirement of building an AI learning dataset. Both API and on-premise are available.
Return Zero (VITO)
Korean AI startup Return Zero offers a voice recognition app called Vito. Vito provides phone call transcriptions in text messaging form. It is perfect for those needing to take notes during calls or transcribe their calls afterward. To date, Vito has processed over 2.5 million hours worth of speech. Vito’s recognition accuracy for unclean speech data, referring to general phone quality, stands at around 80%. As for clean speech data such as audiobooks, Vito’s recognition accuracy is close to 100%. Their aim is to offer the best level of Korean speech recognition worldwide and an enterprise service by the end of 2022.
Posicube
Posicube is an AI-powered voice recognition startup that offers two services. They have the Robi Reception, which was developed to replace repetitive tasks from call centers. The other is Robi Store, used in restaurants and beauty stores to help customers make a reservation. Their AI platform can respond using their data-processing technology in less than a second. Moreover, it can understand customers’ voices more accurately. Since there are no technological hurdles when offering AI services in different languages, Posicube plans to go global in 2022.
Mycelebs
Mycelebs is an AI startup in Korea that offers a search solution called Key Talk that also recommends products that suit a user’s taste. Therefore, their AI technology analyzes the public’s taste through data. Korean companies such as Kakao, Yanolja, Shikshin, Lotte Group, Shinsegae Duty-Free, and more than 30 global companies, including Yelp and travel platform Booking.com, are their customers. Adding Key Talk to a voice recognition engine such as Siri increases the voice recognition rate from 30% to 99%. Their aim in 2022 is to enter the global market and become a global startup that helps various companies transition to AI-based services.
Humelo
Humelo specializes in AI-based voice synthesis solutions. Their text-to-speech solution, AI TTS, can create an AI voice that can read the text to form a short audio recording. These voices are like human voices as they can express emotions and the correct intonation. Currently, the technology requires a sample recording of at least 30 to 60 minutes; however, Humelo only needs a sample recording as short as 2 minutes. In addition, Humelo’s technology can convert the Korean language into other languages such as English and Chinese. Humelo raised $2.6 million from KT Investment and Kakao Investment.
Atlas Labs (Switch)
Atlas Labs is the creator of Switch, an app that allows for real-time transcription using AI technology. This means there is no longer a need to take notes. All phone conversations can be transcribed. The transcriptions are provided in a text messaging format. In addition, users can tap the text bubble to get the audio. The speech recognition solution is called Zeroth, and it can provide the speech-to-text service on the Switch App. Best of all, all the call data is managed by an encryption key that is only shared with the actual user. This means that third parties will not have access to the data.
The app is free to use, and users get 30 minutes of transcription for outgoing calls and unlimited incoming calls. For $9 a month, they can get unlimited transcription services for incoming and outgoing calls.
