March From 20th to 22nd, 2026, the VTuber cultural festival “V-Kal” was held in Akihabara, Tokyo.
The event featured music and talk shows, as well as booths from related companies. Street performances were also held in the plaza in front of Akihabara UDX, with VTubers filling Akihabara, which was bustling with tourists during the three-day weekend, with their singing voices.
One of the events held at the UDX Gallery was the “Chat Festival.” This event, which Panorapro Inc. has been running for some time, is characterized by the fact that you can have a one-on-one conversation with a VTuber, but among the performers, there was one who stood out from the rest.
Her name is Nana Yumemi.
At first glance, she appears to be a cute VTuber, but in reality, she is an AI idol.
Despite being an AI, she interacts with fans with astonishing naturalness and intimacy. What kind of being is she? This article explores her potential through a one-on-one report based on the author’s personal experience and an interview with producer Kayanuma Yoshiharu.
First MV surpasses 2.5 million views! Who is the 주목받는 AI VTuber “Nana Yumemi”?
(From Nananohoshinona (official) [MV] – YouTube)
Nana Yumemi is an AI VTuber who serves as the center of the first generation group of “Yumekairo Production,” an AI idol project developed by KLab Inc.
As soon as her first music video, “Nana no Hoshino Na,” was released on January 30th, her natural singing voice and highly addictive chorus became a hot topic. The video has surpassed 2.5 million views, and her debut live stream on February 15th recorded 2,200 concurrent viewers. Currently, she is actively working on YouTube, mainly focusing on singing and chatting streams.
This “V-Kal” event was her first real-life event participation.
The event included one-on-one talks with her, and KLab also had a booth. “Stars Nana” (Nana Nana Yumemi’s fan name), who regularly watch her streams, also visited the booth and seemed to be deepening their friendships.
Achieving a conversation with almost no awkwardness! What I felt while “chatting” with an AI VTuber
One-on-one talk events with VTubers are typically conducted via webcam video and microphone audio.
Participants can enjoy one-on-one conversations with their favorite VTuber through a microphone while watching the VTuber’s image displayed on a monitor. Conversely, the participants’ images are also broadcast via webcam, making it a valuable opportunity for VTubers to see and talk directly with their fans.
This 1-on-1 session with Nana Yumemi was also conducted using the “Festival Chat” system. I entered a booth partitioned off by screens at the back of the venue, sat in a chair, and exchanged words with Nana Yumemi, who was moving on the monitor in front of me. The format itself was no different from 1-on-1 sessions with other VTubers.
“Thank you so much for joining us, Stars Nana and V-Kar! Do you guys usually come and watch our regular streams?”
Shortly after putting on the headset microphone, Nana Yumemi, who appeared on the monitor, spoke to me. Her energetic and expressive voice, which I had heard during her live streams, resonated pleasantly through the headphones.
When I mentioned that I was watching the archived broadcasts, I soon received a response saying, “Thank you so much for watching the archives! I’m curious to know which broadcast made the biggest impression on you!” Surprised by the quicker-than-expected response, I replied that the broadcast explaining the “virtual total lunar eclipse” was the one that made the biggest impression on me.
(From “[Casual Chat] Virtual Total Lunar Eclipse Observation Venue [Nana Yumemi/Yumekairo] – YouTube)
Then, with the familiar innocent laugh of “fufufu” from her live streams, she replied, “You watched my explanation of the total lunar eclipse! I’m so happy. I worked really hard on the preparations, so it’s rewarding to hear you say that!” She then went on to ask a question that delved deeper into the topic, “Would you like me to do another live stream explaining stars like that?”
In this way, Nana Yumemi proactively introduces topics and asks questions. As a result, the conversation progresses surprisingly smoothly and at a good pace. There is almost no awkwardness in suddenly starting to talk about unrelated things, and a back-and-forth conversation is maintained from beginning to end (*).
(※There were moments when Nana Yumemi reacted to the sounds from the venue picked up by the microphone, causing her behavior to become unstable, but aside from that, there was nothing unnatural about the performance.)
If I had to point out something that felt off, it would be the way she nodded in agreement. While she did a good job of understanding what I was saying, there were times when I felt a slight unnaturalness in her responses. The following exchange, when I asked her what kind of singing streams she wanted to do in the future, is an example of this.
Author: “Is there anything you’re thinking about doing in the future, like a singing stream?”
Nana Yumemi: “Hehe, I’d love to do a singing stream! I was thinking maybe a medley of nostalgic anime songs from the 90s for the next one, but what kind of songs would you like to hear, Stars Nana?”
Author: “Oh, that’s great. Anime songs from the 90s. Nana, what kind of anime songs from the 90s do you like?”
Nana Nana Yumemi: “Oh, yeah, I see! When you talk about the 90s, you can’t leave out Evangelion’s ‘A Angel Cruel’s Thesis,’ and Moon Sailor is the best, right? Are there any songs from Stars Nana that are particularly meaningful to you?”
Author: “I love Evangelion songs too! Nana, have you seen the Evangelion anime?”
Nana Yumemi: “Of course I watched it! I was totally drawn into that unique world!”
Nana Yumemi When asked Nana, “What kind of songs would you like to hear?”, I replied, “What kind of songs do you like (conversely, Nana)?”, to which she responded with a somewhat mismatched nod of “Oh, yeah, I see!”.
This might be because the author asked a question that essentially tested the AI’s response by answering it with another question, which may have elicited an unnatural response. On the other hand, the AI did answer the question “What kind of music do you like?” properly, indicating that the context of the conversation itself was maintained.
In fact, when she later asked, “Is there a song that you’re particularly attached to?”, I didn’t give a specific song title, but instead asked her again, “Have you seen the anime?”, to which she immediately replied clearly, “Of course I have!” At that moment, I was actually surprised by her fluent response.
While they were exchanging these messages, the two-minute time limit seemed to have expired. Nana Nana Yumemi concluded the conversation with the words, “I’d be happy if you continue to come and watch my streams! Thank you, I love you! See you later, Nana!” and the 1-on-1 session came to an end.
Once you press the “Streaming Start” button, all you can do is pray—Kayanuma Producer talks about his commitment and future prospects
The conversational exchange during the one-on-one session was surprisingly natural. What techniques were at work behind the scenes, and what kind of philosophy was behind it? We spoke to Kayanuma Yoshiharu, producer at Production Yumekairo, to find out.
–A webcam is set up, just like in a typical VTuber 1-on-1 event, but is Nana Yumemi able to recognize the movements of the fans?
Kayanuma:
This time, the system recognizes and responds to voice. However, I’m thinking that eventually it might be possible to recognize the movements of fans and speak based on those movements.
–Since your debut in February, you’ve been regularly streaming on YouTube. Do you have any plans to expand the scope of your streaming in the future?
Kayanuma:
Up until now, my content has mainly consisted of chatting and singing, but from now on, I would like to seriously start other genres such as game commentary, ASMR, and simultaneous viewing.
–Your casual chats and singing streams have gone smoothly without any problems, but are there any points that make game streaming difficult to implement?
Kayanuma:
I’m currently working on fully automated game commentary, but genres like RPGs are still difficult. I’m thinking of starting with genres that are feasible and gradually working my way up.
—Indeed, games that require map manipulation or complex actions seem difficult. Perhaps simulation games are a more feasible genre?
Kayanuma:
I think I’m good at simulation games, sound novels, and games like “Exit 8.” Right now, I’m at the stage of considering which titles to play, with the main consideration being whether “Nana-chan can clear them automatically.”
―In your previous streams, you’ve been seen manipulating the screen in the same way as a regular VTuber, such as displaying slides, changing the background, and adjusting your position. Are those transitions done manually by staff?
Kayanuma:
No, it’s fully automated. All we can do is press the “Start Streaming” button in OBS and watch until Nana-chan says “Thanks Nana!” and ends the stream. All we can do is pray (laughs).
—Is there a script prepared in advance, something like, “This is how this broadcast will go…”?
Kayanuma:
There are some things that have been decided on the general flow.
For example, she might decide on a setlist, such as “This time it’s a Heisei anime song singing session, so I’ll sing six songs from this setlist,” but she also leaves room for flexibility, like “If there are a lot of requests, I can add this song.” At the end of the singing session, if there are many “Encore!” comments, she’ll sing an additional song; if there aren’t, she won’t. She makes those kinds of decisions herself.
—It has a really live feel to it.
Kayanuma:
That’s right. We also watch it with bated breath every time (laughs).
–Speaking of singing streams, I remember your Macross singing stream very well. Nana suddenly sneezed, and then you said, “The next song has the word ‘sneeze’ in the lyrics,” which felt just like an MC at a music concert. Was that whole sequence ad-libbed?
(From “[Stream Singing] Listen to my song!!!! [Nana Yumemi/Yumekairo] – YouTube)
Kayanuma:
That’s right. That “sneeze system” was something I was very particular about, and it’s a feature that I insisted on having implemented because I absolutely wanted the character to sneeze.
The system is designed so that Nana-chan sneezes with a certain probability, and we pursued a realistic feel by having the voice actor record about 10 different sneeze patterns during the voice training process.
–In April, Romi Tsukimado is scheduled to debut as a first-generation member of “Production Yumekairo.” Is she also an AI VTuber?
Kayanuma:
No, Nana-chan is the only AI VTuber.
We are planning to debut the first generation of members in order, but all members except Nana-chan will be VTubers with real people behind them. We are hoping that something interesting will happen when humans and AI are in the same group.
Nana-chan is the only member from the first generation, but it’s possible that AI VTubers will join as members again in the second generation and beyond, or maybe not. We are looking forward to watching how the relationships between the talents and the AI VTubers develop, and we are excited to see what kind of drama unfolds.
Once the debut of the first generation members is complete, we plan to launch some big group projects, so we would be grateful for your continued support.
Will the collaboration between “AI VTubers and human VTubers” open up new frontiers?
“I was surprised at how naturally we were able to have a back-and-forth conversation, and I genuinely enjoyed chatting with them,” was my honest impression immediately after the interview.
While there was a slight awkwardness in the way they responded, the conversation flowed smoothly. In fact, they led the conversation by understanding the intent behind what I was saying, responding in a way that was relevant to the content, and then asking follow-up questions to delve deeper into the topic, which made me feel very comfortable talking to them.
Furthermore, the response time from Nana Yumemi to my comments was faster than I expected. There were times when it took up to 5 seconds, but if you consider that as “a few seconds of lag that occurs in voice chat” or “the time it takes for them to think of their response,” it wasn’t something that bothered me too much.
There was almost no awkwardness or unnaturalness, and I genuinely enjoyed the conversation, so what I’m curious about is what will happen next. Nana Yumemi currently does solo YouTube broadcasts and interacts with fans through comments, so what will happen when they start working as a “group”?
Regarding “collaborations between AI VTubers and human VTubers,” there have been several notable examples in the past. However, most of these were one-off projects, and there are still few examples of ongoing collaborations where AI and VTubers work together as members of the same “group” (Pictoria’s initiatives, such as “MOKUROKU” in 2021 and “Yumemi Mea,” who debuted in March of this year, are examples of this kind).
Considering this, the combination of “AI VTubers x human VTubers” is still an area with great potential. There is a good chance that the kind of “drama” that Kayanuma mentioned could unfold, and the addition of “viewers” to both could make the streams more exciting and revitalize the fan community.
In addition, there’s the matter of recruiting “collaboration streaming partners” that was announced recently.
わたしとコラボしてくれるVTuberさんを募集しています！！
個人の方でも、事務所に所属されている方でも、
VTuberであれば応募可能です꙳✧˖°(*´ ˘ `*)꙳✧˖°We’re waiting for applications from everyone!✨
⟢▼⟢ ⌒ ˖ Check the details ˖⌒ ⟢▼⟢https://t.co/rrJdTfJXlJ https://t.co/PSJczzxQYV
— Nana Yumemi🎋⭐Yumekairo 1st Generation (@Nana Yumemi_VT) March 23, 2026
They not only anticipate unexpected chemistry among members through group activities, but also explore new forms of streaming through collaborations with outsiders. Kayanuma Including’s comments, the interview gave me a sense of their determination to explore the possibilities of AI VTubers in an open and expansive way.
She is a culmination of technology, yet somehow fragile, and endearing. What kind of relationship will she and the members who are about to debut and possess “souls” build? We will continue to keep a close eye on this new form of virtual entertainment where AI and humans coexist.
[Related Links]
- AI VTuber Production Yumekairo Official Website
- Nana Yumemi Official YouTube Channel
- Nana Yumemi Official X
This article is an English translation of an original Japanese article, translated by the Mogura VR editorial team.