The Doubao App Has Been Updated with Realtime Voice Call Feature

On January 20th, the Doubao app updated its real-time voice call feature, which is now available to all users.

This feature is based on the latest Doubao Realtime Voice Model. After the update, the conversational ability in Chinese scenarios within Doubao has achieved nearly “indistinguishable between human and machine” AI interaction effects in terms of voice realism and emotional expression of “joy, anger, sorrow, and happiness”. It can mimic different voices and has shown significant improvements in “logical thinking” and “emotional perception”.

In terms of product performance, the Doubao App has achieved a “human-machine indistinguishable” level of real-time voice call interaction, with a qualitative improvement in its voice performance and anthropomorphism. Compared to most voice systems that still make rough changes at the intonation level, the new real-time voice call feature of Doubao can accurately control details such as rhythm, childlike tone, volume, breath sounds based on the scene automatically, and even whisper to you.

Furthermore, Doubao’s expression of emotions such as joy and sorrow is also quite remarkable. It has mastered some dialects for conversations in English dialogue, multi-role imitation, and even partial singing abilities. In daily use, it can be both an English practice teacher and a skilled storyteller or an impromptu songwriter.

In the past, traditional speech dialogue systems used a cascade mode of ASR+LLM+TTS, which could not meet the requirements of completeness in understanding, naturalness in generation, low latency in interaction and other dimensions for human-level voice dialogue. The new voice capabilities of Doubao are based on an innovative end-to-end framework that deeply integrates speech and text modalities using native methods for unified modeling. Ultimately, it can achieve direct multimodal input to multimodal output effects, giving AI voice dialogues a ‘soul’.

According to the relevant person in charge at Doubao, in terms of delivery experience, Doubao’s voice dialogue ensures that the model has strong comprehension and logical abilities to answer real-time questions while also having ultra-low latency and smooth interruption capabilities.

Doubao’s new real-time voice call function sets it apart from similar products with a significant lead in Chinese conversation quality as well as high emotional intelligence and IQ online. According to external feedback, users’ overall satisfaction with Doubao’s newly launched voice call feature is rated at 4.36/5 compared to GPT-4o’s voice dialogue satisfaction rating of 3.18/5. In particular, Doubao has clear advantages in terms of naturalness of tone and emotional richness.

SEE ALSO: ByteDance’s Doubao Launches AI Programming Feature

Report