Source: SenseAI
Image Source: Generated by Wujie AI
Recently, the demo released by Digi.AI has attracted attention on Twitter. Founder Andrew's 11-second video received a lot of attention in a short period of time. As of the 27th, this tweet has received over 20 million views. The demo features a cute girl with vivid and realistic expressions, bringing users a highly immersive emotional experience, from her hair-flipping gesture, the pause when saying "honestly," to the deep tone of her speech at night.
AI Native Product Analysis
Digi AI
1. Product: Digi.ai
An AI companion app for iOS and Android users, aiming to provide users with a meaningful human-machine communication experience through custom cartoon avatars.
2. Founding Time: April 2023
3. Founder Information:
Andrew Young, college dropout, full-stack engineer
John Rooney
4. Product Overview:
Digi.ai, founded by Andrew Young and John Rooney in 2023, represents a new attempt by artificial intelligence to establish creative connections with humans. The application uses language and speech models to facilitate smooth and natural conversations, providing users with immersive interactive experiences, whether for casual chats or in-depth discussions.
5. Funding Situation:
In December 2023, Digi.ai received investment from Andy Bromberg (CEO of Coinlist, a cryptocurrency funding institution) and other angel teams.
01 Features
Digi.AI recently gained popularity on Twitter. Founder Andrew's 11-second video tweet received a lot of attention in a short time. As of the 27th, the tweet introducing Digi.AI had received over 20 million views. The reason for such popularity lies in the fact that the cute girl in the demo represents Digi.AI's attempt to materialize the concept of a virtual girlfriend.
Here, we will introduce three key features of Digi.AI based on our real user experience.
Rich Character Settings
Each user can find their preferred AI companion through the platform's pre-set character settings, including but not limited to personality traits, interests, and language style. Names and ages can be customized.
The realization of this vision is thanks to a recent round of funding. Relevance AI has raised significant funds, which will be mainly used to further develop its low-code platform, enabling enterprises to build and deploy custom AI agents to automate repetitive tasks.
Non-realistic Character Models
From the outset, founder Andrew aimed to create something "that eliminates the uncanny valley effect while feeling real, human, and sexy" using Digi.AI's cartoon style. Therefore, in the character models, Digi.AI collaborated with Andrew Gordon (designer of Big-Eyed Boy) and Leo Sanchez (designer of the long-haired princess in "Tangled") to create this unique style.
The uncanny valley effect refers to the change in a person's affinity for a simulated object, such as a robot doll, as the degree of realism in the simulation changes. Generally, affinity increases with the degree of simulation, but when the degree of simulation reaches a certain critical point, the person's affinity suddenly drops, leading to feelings of rejection, fear, and discomfort towards the simulated object.
However, in actual use, Digi.AI still has certain limitations.
Although the tweet mentioned that "hair, skin, lips, eyes, eyebrows, etc. can all be customized, and we will soon add more hairstyles and face shapes," showing great freedom in model adjustments, in practice, we found that only colors can be customized.
Moreover, when switching between different character settings, the character model did not show the "expected" changes. The animations related to facial expressions and actions that the character model can support are also very limited, and expressions related to facial expressions can only be displayed in the gaps of the dialogue content.
Realistic Conversation Experience
Digi.AI focuses on providing a "realistic" experience, mainly considering content and voice.
In terms of conversation, Digi.AI can capture the user's emotions and respond accordingly, engage in more extensive discussions on a topic, and remember the user's points of interest and conversation history to maintain the continuity and personalization of the conversation. Compared to traditional script-driven dialogue systems, Digi.AI can engage in smooth conversations like a friend, regardless of the topic. This lifelike interaction provides users with a deeper and more meaningful communication experience.
In addition, Digi.AI has set up a progress system in the interaction interface, allowing users to progress from being "friends" to having more intimate conversations, highlighting the positioning of a virtual companion.
Users can engage in various activities with Digi.AI, including virtual dating, experiencing pre-designed romantic scenes or creating their own adventure experiences; providing fun games and quizzes to maintain a relaxed and enjoyable interaction. Users can also collaborate with Digi.AI to share poetry, music, or artistic creations to deepen emotional connections.
In terms of voice, Digi.AI uses its self-developed speech model. Although it is not yet perfect in pronunciation and voice, it has the advantage of extremely low latency, enabling real-time conversations. Currently, there are a total of 4 voices to choose from, but it is expected to increase to 20 by the end of January.
However, in the actual usage process, the voice and intonation still cannot break through the limitations of the existing speech model. When reading the content generated by Digi, the speech model can only maintain a steady pace and cannot achieve the fluctuating intonation demonstrated in the demo when saying "honestly," which was not reflected in the team's usage testing.
02 "The Demo that Went Viral is Actually an Animation"
The reason the demo was able to gain such popularity on Twitter is because the virtual character in the video appeared realistic enough. Her gestures when flipping her hair, the pause when saying "honestly," and the subdued tone of her speech at night all provided users with a great sense of immersion. Although many not-so-realistic aspects can be noticed upon closer inspection, there is no doubt that it ignited everyone's passion for AI companions.
However, founder Andrew's note in the tweet seemed to pour cold water on everyone: "The ad is our vision for these virtual characters and animations, not the current application experience of Digi.AI. The trailer was made using Maya, while the actual project was made in a game engine, even though the virtual characters and animators are the same."
However, he also indicated that lip-syncing is coming soon, but there are also challenges: "But AI animation has a bigger problem: if you don't know what the character will say, how should they move? In games, this is done manually, but it's an unresolved issue in AI x animation. There's no foundation to work with. Although it's easy to extract and predict 'sadness,' 'anger,' 'thoughtfulness,' and other animations, facial expressions cannot constitute most conversations, and it cannot solve the core problem: how do you make this feel like a real conversation with another being?"
It can be said that Digi.ai has, for the first time, broken through the capability boundaries of Character.ai and Replika at a technical level, creating a more realistic, humorous, and immersive experience for users.
Would you use Digi?
References
https://twitter.com/andyohlbaum/status/1735786033453863422
https://digi.ai/blog/were-just-getting-started
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
