DubGuild, Inc. (Headquarters: Bunkyo-ku, Tokyo; CEO: Masatoshi Otake), a company engaged in the R&D of voice-interactive AI foundation models, is pleased to announce its participation in the 'AI Foundation for Startups (AIFS)' program provided by SoftBank Corp. Through this program, we will accelerate the advancement and business deployment of our audio-only trained interactive foundation models.

Background of AIFS Participation In recent years, the evolution of generative AI has brought increased attention to the voice AI field. However, conventional voice AI has largely relied on a structure of converting audio to text, processing it as language, and then generating audio again. This structure makes it difficult to handle temporal and emotional information such as backchanneling, interaction timing, overlapping speech, emotional nuances, and natural 'pauses' in conversation, which are often lost during the text-conversion process.

Technical Features of DubGuild Unlike traditional models that rely on text, DubGuild is developing a foundation model that learns directly from audio. This approach significantly increases the amount of information the AI can handle, enabling natural backchanneling, processing of overlapping speech, audio generation reflecting emotional expression, low-latency real-time translation, and interactions that maintain the temporal structure of conversation. This is not merely voice recognition technology, but a voice-specialized foundation model that handles the very structure of human conversation.

Future Outlook We will leverage the large-scale GPU computing infrastructure provided through AIFS to further advance our voice-specialized foundation model. Additionally, we will strengthen collaborations with companies and research institutions to accelerate pilot experiments and commercialization. We aim to expand globally in the fields of AI voice dubbing, multilingual real-time translation, and next-generation voice interaction.

CEO Comment CEO Masatoshi Otake: 'We are serious about breaking down language barriers. There is a wealth of wonderful content in the world today, but much of it is blocked by language barriers and does not receive the recognition it deserves. DubGuild aims to deliver all local content to the world with voice foundation technology that achieves "dubbing that preserves the worldview." With the tailwind of participating in AIFS, we will accelerate the evolution of our foundation model that understands audio directly and create value that transcends language.'

FACT BOX

  • Source: PR TIMES
  • Category: partnership