Jun Sasaki / Log

2026-04-01

Early YouTube is indeed full of ups and downs. Rt2112 Development Log | March 28 - April 1, 2026

Well then, as the title suggests, after getting it spread around, dropping the CTR, and building up a peak to where it is now, we've hit a valley again.
Now, let's talk about that valley.


Latest channel data (as of April 1st)

  • Main channel: Approx. 25 videos published, scheduled through 4/6

  • Lo-fi 2112: 3 videos published

  • 28-day views: 1,100 (>999% increase)

  • 28-day impressions: 27,200 (>999% increase)

  • CTR: 2.6%

  • Average view duration: 3:38

  • Top videos: 149 views, 82 views, 85 views

  • Browse features traffic: 84% (algorithm continues to push)


Finalizing title tags (5 in description)

  • #SleepStories #BedtimeStories #CozySciFi #GentleFuture #RelaxingAmbient

  • Added sleep stories, bedtime stories for adults, gentle english stories, calm english narration, etc., to the video tags section

  • Slipped #StudyBGM #FocusMusic into the description to expand search reach

  • Policy of not narrowing down by age or gender in tags (do not include 'adults' in title tags)


Positive factor: 84% browse features

  • 84% browse features traffic by the 15th video is an exceptional start

  • The sleep genre has the structure that YouTube's AI likes the most.

  • A 58-minute video generates 5-6 times more session time than other genres.

  • The AI learns the nightly viewing routine of the sleep-deprived audience -> recommendations become a reserved seat.

  • Peaks and valleys in impressions are a YouTube specification. The correct approach is to accumulate them steadily.

👉 The only positive factor. Once released, it will appear in browsing, so impressions will increase. The negative part is where the impressions are being sent.
Since it is scattered across many directions and countries, it is natural that the CTR would drop.

Both a negative and a future positive factor: Auto-dubbing.

If you listen to the Japanese audio, as you might guess, it is a machine voice that is not at a level suitable for bedtime.
While impressions are increasing in places like Ukraine and South Korea, these are not major language markets for YouTube, just like Japanese, so I assume they can tell it is a machine voice while listening. (Though I think it is slightly better than Japanese.)
If so, low watch time is within expectations.
However, if both CTR and watch time drop, it will hit a ceiling, so the next countermeasure is also necessary.

Even in such a situation, there are positive factors.
Languages for Expressive dubbing have been steadily released.
Spanish, Portuguese, German, French, Italian, Hindi, and Indonesian.
The accuracy is high, and I think they are likely much easier to listen to compared to the aforementioned Japanese, Korean, etc.
Therefore, I am moving to the next phase of hypothesis verification.

This is the current hypothesis.

  • The auto-dubbing full language release in 14 days is almost certainly correct.

  • There is some audio distortion in the BGM, but it's not unpleasant, so it should be competitive enough even in local languages.

  • YouTube is prone to bonus impressions in the first 48 hours.

Based on these three hypotheses, here is the verification plan.

  • An operation to secure a 14-day buffer through daily posting scheduling.

Smoothing out the schedule with 3 posts per day.
(Even including days when I'm not motivated) I will definitely accumulate 14 days' worth of stock in one week. By publishing on a 14-day delay schedule, I can ensure initial momentum in all languages, which will increase the expectation for impressions.

I feel like I could just set it to 14 days from now, but since I also have the goal of having the algorithm learn "Bedtime Stories" by posting at the same time every day, I'm weighing both options and...

Going with a brute-force posting approach.

I'll handle it that way.
If I can just keep pooling content in April and schedule about a month's worth of posts, it seems like I'll have quite a bit of breathing room.

Posting pace

  • Publish 1 video per day + stock 3 scheduled posts per day

  • Scheduled through 4/6

  • After securing a 14-day buffer, aim for a 1-month scheduling lead.

  • Goal of 1 post per day (365 videos) for one year.


The day/night issue with thumbnails

👉 Fine-tuning. Previously, I was creating loop animations that matched the story, but...

  • Even though it's a Bedtime title, a daytime thumbnail is weak for first-time viewers.

  • The data also shows that night thumbnails (149 views, 85 views) overwhelmingly outperform daytime thumbnails (6 views, 8 views).

  • All future new releases will be unified to night scenes.

  • Added "Time of day: Evening to late night" to the image prompt

  • Image is night -> Gemini ambient sound extraction is also night-leaning -> Suno BGM is also night-leaning -> Everything clicks

  • Keep existing daytime videos as A/B test data

👉 When searching for "Bedtime stories," most thumbnails are at night, so that makes sense. I'll leave the loop animations I'm currently making as they are and just darken the thumbnails. From now on, I'll be mindful of the time of day when generating animations.


KPI Settings

  • Level 1: 1,000 views on a single video (proof of algorithm recognition)

  • Level 2: 10,000 views on a single video (the line where monetization comes into view)

  • Level 3: 100,000 views on a single video (the line where 500,000 yen per month becomes a reality)

  • The channel as a whole has already surpassed Level 1 with 1,100 views per month

Impression Goals

  • Back-calculated assuming a 3% CTR

  • Level 1 (1,000 views): 33,000 impressions (at 27,200 this month, almost there)

  • Level 2 (10,000 views): 333,000 impressions

  • Level 3 (100,000 views): 3,330,000 impressions

  • Optimistic scenario for 3 months from now is 3.3 million impressions (3x monthly growth + full auto-dubbing activation)

Market size of the Boring History segment

  • The segment that "wants to listen to a soothing voice while falling asleep" is in the millions

  • There is a case of one video reaching 2.3 million views

  • It's currently flooded with AI slop, but there's a high possibility of it being subject to bans

  • rt2112 will serve as a receptacle for original content

June YouTube Update

  • Possibility of auto-dubbing improvements coming

  • Hoping for a resolution to the BGM stuttering issue

  • If I have over 50 videos stocked by then, I can maximize the benefits

Discovery of elderly thumbnails

  • Elderly thumbnails perform better in CTR, view counts, and retention than boys/women

  • Since the audience is skewed toward the 55-64 age group, characters close to their own age resonate

  • Anime-style old man x robot is a point of differentiation no one else is doing

  • Intentionally increase elderly characters in future story production

Formulation of 3-month strategy document

  • Documented the strategy for April to June 2026

  • Includes KPIs, channel structure, production flow, 3 revenue scenarios, milestones, and decision criteria

  • Reviewing the results in 3 months

Next steps

  • Produce 3 videos per day until 14 days of scheduling are complete

  • Add night specification to image prompts to unify all videos to night scenes

  • Additional release of Lo-fi 2112

  • Keep stacking them calmly, aiming to reach 50 videos

Advertisement