Jun Sasaki / Log

2026-04-07

Rt2112 Development Log | April 4 - April 6, 2026

I am creating a BGM channel on the sub-channel side. It is a form of reusing music from the main side.
I am also aiming for people to listen to music when they don't want to hear readings, or to have music flow from the readings.
Above all, managing music is fun lol.

With that said, I am doing this because I want to, so I am uploading on a whim, but since I am doing it, I think I should set some goals.
When I roughly researched the competition, successful channels generally gained 5 million views every six months, while there were also many channels with around 100 videos that could not be monetized and had garbage-like view counts.
This difference is, without needing to investigate, whether or not there is a theme. For example, mass-producing harmless music like what you hear on early morning radio seems to be flagged as spam.
Well, since I only have about 10 videos, I am on that side too, so I will just calmly keep adding to it.
In addition, the English-speaking market is too much of a red ocean, so I added a slight change this time.


Multilingual Title Automation (localize.py)

Title localization using the YouTube API.
I think I could finish it in an hour even if I just typed it all manually, but even that is a hassle, so I will handle it with a single command.


  • Put the previously created localize.py into production

  • Consolidated authentication files into the config/ directory. Resolved the clutter directly under fix

  • Dry run with Lo-fi 2112's Distant City, Pre-Dawn Study video → 25 languages, zero over 100 characters

  • Production execution → 25 language titles set in 10 seconds

  • Visually confirmed Japanese title reflection in YouTube Studio

  • Doing it manually takes 1 hour per video. That is now 10 seconds. Zero cost

  • Title structure: [Catch] [Emoji] Lo-fi [Usage] | Lo-fi 2112 | Distant [Location Name]

  • Usage words are SEO-optimized for each language (JA=Work BGM, ES=para Estudiar, DE=zum Lernen)

  • Established operational flow: Copy existing JSON → Translate only the catch with GPT → Single command

  • Operational guide also created. Now I can run this by myself

👉 Hypothetically, I thought this would make search results easier to reach in niche countries and languages. Impressions have actually gone up since the first day, so this is likely a success.
Currently, I've localized into 26 languages, including niche Asian regions like Tagalog, and Eastern European regions like Czech and Hungarian—where English might suffice, but using the local language makes the search more niche.
Since music has no borders, I believe just replacing the titles is sufficient.

I think translating the description fields would make them even easier to find in search, but I decided against it because I place a mention of the main channel (@Roadto2112) in the first line, and localizing it would remove that mention.
The BGM channel is strictly a sub-channel; the main focus is the bedtime stories.

Main Channel Data (as of April 6)

  • 28 days: Views 1.2K / Impressions 32K / CTR 2.5% / AVD 2:54

  • Compared to previous 28 days: Views +753%, Impressions +821%

  • Algorithm test distribution is ongoing. It is giving me 3,000 impressions at a time.

  • The 2.5% CTR has not yet reached the growth threshold (4-5%). However, this is not the phase to be making weird adjustments.

  • My hypothesis is that the low AVD of 2:54 is caused by the standard TTS quality of the auto-dubbing.

  • The Expressive Speech regions (ES, PT, DE, FR, IT, HI, ID) have listenable quality. The standard TTS regions (JA, KO, UK, etc.) are tough.

  • Country-specific data won't appear until it exceeds 100 views, so this remains a hypothesis.

👉 The bottleneck is that the Average View Duration (AVD) is extremely low for a 58-minute video, but I figure this is a necessary upfront investment. Expressive Speech is quite good, but it takes time, so I am releasing videos based on algorithm priority before the auto-dubbing for those is complete. I feel like people in standard TTS regions like Ukraine and Korea are listening and getting angry, thinking, 'Ugh, I can't sleep with a voice like this, you loser!'
However, since YouTube seems to be investing heavily in auto-dubbing, I have high hopes for the June update. My strategy until then is to keep building up the number of videos.

Lo-fi 2112 Data (as of April 6)

  • 28 days: Views 131 / Impressions 3,801 / CTR 1.8% / AVD 16:16

  • An AVD of 16:16 is about 25% for a 60-70 minute video. That is sufficient for the early stages of a BGM channel.

  • Signs of impressions jumping after introducing multilingual titles.

  • Cozy Cosmic Vibes is the top performer with 42 views. It is the closest to the 100-view line.

  • The difference from the main channel (AVD 2:54) is clear. There is a possibility that the Lo-fi side will reach monetization first.

👉 This side has entered a phase where building up content leads to growth. The fact that impressions are rising is a good sign. For now, I think it's just a matter of building it up.

Google Vids × Veo 3.1

  • Veo 3.1 has been integrated into Google Vids. I'm hoping this means the loop video gacha will become unlimited.

  • Result: Monthly subscription system. 10 free videos/month, 1,000 videos/month with AI Ultra ($250/month)

  • The current 3-gacha-per-day (90 times/month, free) is more cost-effective. Passing on this.

  • OpenAI's withdrawal from Sora. Google is going all-in on video generation. Waiting for the possibility of an expanded free tier.

→ There's something like Canva called Vids on Google, and I saw on X that you could use Veo as much as you want, so I tried it, but maybe it's only unlimited for students.
Right now, I'm using the video generation available 3 times a day in Workspace's Gemini. Honestly, the bottleneck is that if it fails, I can't make even one video a day, but I don't want to spend that much budget, so I'm prioritizing prompts that guarantee success for those 3 shots.
Either way, even on my best days, 3 video generations a day is my limit, so that's fine.
If I built a full-spec environment, I could probably make 30 videos or automate it, but platforms, including YouTube, hate that, so I don't recommend it.
What information peddlers call 'video automation' is just a junk mass-production machine, so platforms won't even share impressions with you.

DJ Character Composition Policy

  • Introduce a robot DJ to the Lo-fi 2112 thumbnails

  • Lo-fi Girl uses a 'fixed character.' Lo-fi 2112 will differentiate itself with a 'fixed composition.'

  • Fixed: Front-facing DJ booth composition, Lo-fi 2112 logo position

  • Variable: Player (cat robot, dog robot, bird, human...), background (Distant theme)

  • Can build brand recognition while avoiding the 'boredom' risk of a fixed character

  • Could also become an MD candidate in the future. Currently in progress with the Claude illustration team

👉 A complete change of pace. The current loop animation is perfectly fine, but I wanted to do something silly, so I made it a robot DJ loop animation.
However, given my current business style, there is no fixed character, and in a sense, the lack of a character is a weakness for an IP, so I have a very shallow expectation that this might be able to complement it.

AINOA Connection

  • Plan to naturally blend AINOA-like city visuals into the background of Lo-fi 2112

  • Will not push it from the Lo-fi side. A structure to be recovered when AINOA launches in 2030

  • Do not include AINOA city names in the Distant series name. Just hint at it through background art

→ Personally, I do want to do a long-form IP, but the design of connecting to AINOA (which is currently being prioritized for MD, and since Rt2112 is the main axis, it's barely moving) is just interesting in a way.

ACID 2112 Concept

  • Concept for an up-tempo spin-off channel as a counterpart to Lo-fi

  • The 4 cities of AINOA (Toyo/Seigetsu/Nanen/Hokuka) have good compatibility with music genres

  • Zero-cost seeding by including city names in album and track titles

  • Considering distribution on Spotify/Apple Music via TuneCore Japan

  • Expand horizontally once the multilingual success of Lo-fi 2112 is confirmed

→ Once the Lofi side grows, an uptempo sub-channel could be an option. But well, doing it now would be completely meaningless, so it's just a thought.

Policy

  • Main channel: Reach 50 videos by May. Prepare for the June algorithm update

  • Lo-fi 2112: Apply multilingual titles to other videos as well. Aim for over 100 views

  • Release videos with completed auto-dubbing ahead of schedule. Keep the 6:00 AM (JST 23:00) release time

  • Optimization will come after seeing country-specific data. Now is the phase for building up

Advertisement