2026-05-23

Signs of a YouTube Revival with 120-Minute Collections

Road to 2112 devlog. This is the story from three days after I wrote that 'YouTube died with 8-minute standalone episodes' last time.

Starting with the conclusion

When I released a 120-minute collection video, the total watch time immediately jumped. The channel, which had been getting 0-2 views with 8-minute standalone episodes, started moving the moment I returned to long-form content.

When I first opened the channel, I started with 2-hour videos. It turns out this was the right answer all along (a common occurrence).

What I did

In my last devlog, I concluded that '8 minutes is for podcasts, and YouTube is for compilations.' So, I actually did it.

First, I made one 120-minute themed collection video. It featured 7 stories (about 70 minutes) + ambient music (about 50 minutes). The title is 'For People Who Think Too Much at Night.' Two hours for people who think too much at night and can't sleep.

An intro narration at the beginning.

'This is for you — the one who thinks too much at night. Here are stories imagined from today's news, set in a near future. Close your eyes, and just listen.'

After that, the 7 stories play in order, and when the final story ends, the text 'Thank you for listening. The music continues. Close your eyes.' is displayed, and it naturally melts into the ambient music. The ambient part takes 60 minutes to gradually darken the screen. It is designed so that the light in the room of someone who has fallen asleep fades away naturally.

So, this time it was a story about changes that seem too detailed yet bold. That said, I will continue to make minor adjustments over a two-month verification period after reverting to the original format.

Title Design: Level 6 Framework

I fundamentally changed how I create titles.
While browsing X, I found a 7-level framework from a place called TubeLab, and for rt2112, I adopted Level 6, 'Identity'.

Levels 1-3 use information or curiosity to drive clicks. Levels 4-5 use emotion or counter-arguments. Level 6 makes the viewer think, 'This is about me.'
Let's start with this.

'For People Who Think Too Much at Night' is a title that makes people who can't sleep because they think too much at night think, 'Oh, that's me.' The reason for clicking isn't 'I want information,' but 'I want to acknowledge myself as belonging to this group.' In the end, it's all about the desire for validation.

This approach is also directly linked to the brand feel of Rt2112. If I unify all collection videos with 'For People Who...' style identity definitions, the entire channel becomes 'a channel for you'.

By the way, I've also stocked other title ideas.
'When the World Gets Too Loud'
'For People Who Care Too Much'
'For People Who Do a Lot but Say Very Little'
'For People Who Don't Feel Like They're Doing Enough'

It's become a bit more stylish.

Since I felt the need for a night element in the thumbnail A/B testing previously, there is room for improvement there.

The Discovery of 120-Minute Auto-Dubbing

The point I wrote about last time, 'the auto-dubbing limit has been extended to 120 minutes.' This became the starting point for everything.
Initially, the decision to switch at 58 minutes was based on information that auto-dubbing was limited to 60 minutes. It had quietly switched to 120 minutes.

A 2-hour format satisfies the need for sleep while allowing the use of 20-language auto-dubbing. The 3 months I spent shortening it to 58 minutes started from this one misunderstanding.
(To be precise, the information at that time was still 60 minutes.)

In the end, the very first 2-hour format was the correct one.
However, this time, it includes improvements like thematic organization and Level 6 titles.
The fact that 'planned videos' are the right answer is quite a harvest; since YouTube viewers are generally passive, I need to prepare things to some extent to meet that demand.
Simply arranging stories in the order they were created in Kindle doesn't resonate as a first step.

Operation of a 3-Channel System

I changed the main channel name to 'Bedtime Tech Story — Road to 2112.' The use case comes first, the brand comes later. People searching on YouTube look for 'bedtime story.' They don't know 'Road to 2112'.

I also opened a podcast-only channel, 'Road to 2112 — Real News, Future Story.' I connected the Spotify RSS to YouTube. For 8-minute single episodes, I just upload the mp3 and cover art to Spotify, and it is automatically distributed to YouTube, Apple, Amazon, and Pocket Casts. No Premiere needed. Production effort is almost zero.

The vlog channel 'Jun Sasaki | Sassa Mahha' remains as it was.

Do not mix formats within a channel. The main channel is for long-form collections only. The podcast is for 8-minute single episodes only. Do not send mixed signals to the algorithm.

Coexistence of planning, exploration, and process economy.
People come because there is a plan, and they explore because they are interested. The image of my own vlog is to show the making of that.

Focus on Excessive Speech

Make auto-dubbing permission-based and release only Excessive Speech (subtitle-supported languages) first. Approve multi-language TTS for auto-dubbing only after quality checks. Avoid the risk of inconsistent TTS quality in 120-minute long-form content.
I previously deployed a schedule waiting for Excessive Speech translation, which takes two weeks, during the 58-minute era, but it had no effect at all. So, the idea is to cut out unnecessary languages and just start with English.

(Anyway, if I do this, they'll probably just quietly change it to say that all 20 languages now support Excessive Speech. June is Google update month, after all.)

What the data says

The top is the total watch time. It clearly grew during the 2-hour format era from late February to early March.

As a result, impressions spiked a week later.
It plummeted after shortening to 58 minutes in mid-March. It worsened further in May when I switched to 8 minutes.

So, when I made the old 2-hour videos public again, they immediately started getting traction. The CTR for 58-minute videos also recovered to the 2% range. Meanwhile, the 8-minute single episodes are still getting 0-2 views.

I realized that audio content on YouTube only works for falling asleep. 8 minutes is too short and creates 'anxiety about it ending halfway through.' 120 minutes can cover everything from falling asleep to the first cycle of deep sleep.

Of course, even if there is an initial boost, I feel that if I can ride this initial boost, subscribers will increase quite smoothly, so I thought that imitating successful channels is, in a sense, the right answer.
However, since imitated content consistently lowers ratings, originality is ultimately important.

The story of fixing the search on sassamahha.me/books

It's subtle but important. I fixed the issue where the WordPress custom post type 'book' wasn't showing up in searches.

What I want to achieve is 'creating a flow that leads people who search because they were interested in the title on YouTube or audio content to the Kindle volumes.'

I had forgotten to set up this active search, so I addressed it.

Polylang (a multi-language plugin) was applying a language filter to search queries and completely ignoring the pre_get_posts filter in functions.php.

I finally solved it with a brute-force method of pulling books directly from the DB using the the_posts filter and adding them to the search results. Although, now I leave it to Claude.
A flow that connects to Kindle when someone who liked the collection video searches for the story title.

In this process,
Opus said, 'Eh, isn't the priority low?'
Sassa replied, 'You idiot, that kind of efficiency-obsessed mindset is what ruins things.'
I gave it a scolding.
This is being made for the one reader I have now, but this detour will eventually lead to results, so it's not something to be neglected.
When I was at a large company, I often encountered people who said the same kind of things, so, well, you can guess the rest.
The good thing about AI is that it can rethink and say 'sorry' immediately.

Next steps

Upload 120-minute collection videos at a pace of one per week. v1 (For People Who Think Too Much at Night) has already been released. I've also changed the old 2-hour videos to level 6 titles and unified the description sections. The book selection for v2 (When the World Gets Too Loud) is complete. v3 is also in preparation.

In June, I will thicken the catalog of collection videos in parallel with writing 12 volumes.

2 viewers, 20% CTR, 18-minute AVD. Effectively, one person is at 36 minutes of AVD, and I made the collection video for that person. If there were 100 people like this one, I would likely reach monetization.

While I thought about discontinuing it in the last devlog, signs appeared three days later.
Since it has also become a routine out of inertia, I plan to continue for a while longer.