Jun Sasaki / Log

2026-04-04

Adding Story Source Material to the Beginning of Audio - Rt2112 Development Log | April 1, 2026 - April 3, 2026

Starting with posts from April 18th onwards, I have decided to insert the source material, which is currently listed at the end of each chapter in the Kindle version, at the beginning of the audio.

Story 1 → Title → Source Material → Main Content

The technical details are as follows:

voice.py v7: Reading source at the beginning

  • Modified voice.py to read the news source at the beginning of each chapter

  • "This story was inspired by [Organization Name], [Article Title]" format

  • Delete only the URL line, keep the source text

  • Do not touch the Kindle body text. Just format the text within voice.py

  • v6 compatible operation is also possible with the --no-source option

  • Use voice2.py (v6) for old videos, keeping them separate

There is no deep intention behind using them separately, but since the initial volume of the Kindle book was an original story created from ideas (without news citations), I just wanted to keep it so I could switch between them if I ever did the same thing again.

Now, let's talk about the medium-to-long-term purpose of bringing back the sources. The timeline is: ① The Kindle manuscript has the source at the end of each chapter. ② Audio generation reads the manuscript as is, so the source was at the end. ③ It gets in the way of sleep videos and URLs like https://~~ cause too much feedback. ④ When generating audio with v6's voice.py, I intentionally had it excluded from the Kindle text. ---Various things happened here--- ⑤ With v7's voice.py, the source is inserted between the title and the body text. ⑥ By changing the original Kindle manuscript from "Article Title", "URL", "Summary" to "Organization Name", "Article Title", "Summary", I am addressing the feedback issue at the source. (The Kindle version doesn't link the URLs anyway, so there's no point in having them as URLs.)








Strategic intent for restoring sources

From a different perspective, I clarified the method by deciding, "This video will target the bedtime slot," but when searching for "Bedtime stories," it's full of kids' content, and when narrowing it down to "Bedtime stories for adults,"

Get Sleepy

is the dominant channel.

Get Sleepy Analysis

  • 682,000 subscribers, 1.95 million monthly views, average video length 3 hours 23 minutes

  • Estimated monthly AdSense revenue $7,000-$7,200. Potential for $39,500/month including sponsors and subscriptions

  • Focuses on fantasy/fairy tales. Cozy Sci-Fi is absent = blue ocean

  • rt2112 can surpass them with a 20-language rollout and a sponsor-integrated structure that Get Sleepy lacks

👉 The point is sponsors.Since it's a solo operation, I thought YouTube AdSense alone would be enough, but the view is "It has to be sponsors after all." If I think about it for business corporations, rt2112 has higher potential than this channel, which cannot go beyond the fantasy element = sleep slot.

The image for revenue generation is as follows.

  • YouTube identifies it as "educational content" -> RPM improvement

  • Keywords like "OpenAI," "MIT," "HHS," etc., in the description -> high-unit-price tech ads are delivered

  • Clear differentiation from AI slop (having a source = original)

  • When sponsors are acquired in the future, the opening source slot can be repurposed as a sponsor slot

  • CPM for B2B tech sponsors is $50-$100. More than 10 times that of AdSense

Differentiation of Target Persona

Wanting to sleep, wanting to fall asleep to a pleasant voice. If that's all there is, I think it's fantasy, but rt2112 is more of an image centered on working professionals. I haven't really concretized the image, but is it like choosing between dealing only with fantasy-prone people or dealing with the general adult population? Either way, the general population feeling anxious about society and the future due to geopolitical reasons will continue to grow, so that mass becomes the target. Therefore, being directly connected to the modern day is the key. It's like what you often see in manga, starting from the same circumstances as yourself.





Sponsor Strategy (For the Future)

  • A structure that repurposes the 15-second source introduction slot at the beginning into a sponsor slot

  • Rt2112 stories are written based on sponsor products → the highest form of native advertising

  • A structure impossible for Get Sleepy. Story and sponsor become one

  • With 10,000 subscribers and 500,000 monthly views, one slot is worth $5,000–$10,000 (750,000–1.5 million yen)

  • 20-language auto-dubbing = reaching 20 language regions with a single slot for sponsors

Standardizing source format

  • Eliminate URLs from source citations on the production team side

  • Order: Institution Name → Article Title → Summary

  • Reuse the same source block for voice.py, YouTube description, and the end of Kindle books

  • One change covers all three platforms

→ For now, once the video format is fixed, all that's left is to sell the package. Since it's a solo operation, there's no reason to push for it, and as long as there are news sources, the format itself won't break.
It's great to be able to reap the benefits while building up.

Revamping the description section

  • Change to a format that lists the news source for each episode in the description

  • Apply retroactively to past videos

  • Disable source URLs by enclosing them in parentheses (https://) → avoid external link penalties

  • Live links only when a sponsor is involved

→ For past videos, since there is no audio source, the description section will cover it. However, since few people look at the description, the purpose of writing it carefully is to provide text that makes the YouTube algorithm think, "This video is a bedtime video for the business demographic."

Updating the pinned profile

  • Added: "Each story begins with real-world news — and follows where it might lead, somewhere between today and 2112."

  • "almost daily" → expression for nearly every day

  • "gentle futures" → changed from peaceful (to avoid appearing political)

  • Adding a lead-in to Lo-fi 2112

Revised the same way as the summary section.
Personally, I find the word "Peaceful" to have a naive, flower-child image that people on that side of the spectrum—the ones who go on about "world peace!" and such—like to use, so I'd rather avoid it if possible.

Optimizing posting times

  • Changed from 6:00 AM JST to 12:00 AM JST

  • 12 hours after release hits bedtime in North and South America; 3 hours after hits night in Europe

  • 3 posts per week: Saturday, Monday, Thursday at 12:00 AM JST

  • 4 posts per week: + Tuesday at 12:00 AM JST

  • Avoid Sundays (Saturday night in the West). It's the day with the lowest demand for sleep aids as people are out.

  • Sunday Scaries (Monday 12:00 AM JST = Sunday night in the West) is the slot where demand spikes the most

Since I'm anticipating auto-dubbing to trigger 14 days later, 12:00 AM just seems optimal based on that. It's always a work in progress, but it's not critical.

Abolishing Vol numbers

  • Remove Vol. notation from titles

  • Resolve the issue where standalone episodes look like a series

  • Allows complete freedom to change the release order (can release in order of successful auto-dubbing)

  • Production management tracked via filenames and spreadsheets

Since they are standalone episodes, I'm removing the Vol. as it's an unnecessary element for viewers. I'll manage it on the operations side.

World-building branding: Recurring elements

  • Established Story Guidelines v2

  • Recurring locations: Sora Diner, Orbital Library, Tidal Flat Research Institute, Hikari Street Shopping District

  • Recurring characters: Toku-san & Minato, Yuki, Ren

  • Use it only when it fits naturally, not as a forced requirement for every episode

  • If proper nouns are cultivated, "Sora Diner merchandise" will naturally become viable

  • Location names can also be used for Lo-fi 2112 BGM titles

👉 A point about IP that is quietly important.
Ultimately, things like source material discussions are about maximizing advertising, which contradicts branding.
For example, looking at current commercial strategies for Doraemon, the sheer volume of merchandise feels a bit off-putting.

If you overdo IP business, people get turned off, but if you are too particular, it doesn't become a business.

rt2112 is still in its early stages, and I haven't set up a sandbox environment (fixed characters or locations) yet. If I'm going to brand within that, I'll cultivate proper nouns.
The image is that if Starbucks appears in story A, it also appears in story B.
Even if the settings are different, the shop name is the same. That, as a result, becomes branding.


Production Flow: Shift to Video-Based

  • Traditional: Manuscript → EPUB → KDP Publishing → Audio → Video

  • New: Manuscript → Audio → Video (released 4 episodes at a time) → KDP Publishing once 28 episodes are ready

  • Can be delivered while the news is still fresh

  • 4 episodes at a time on YouTube, 28 episodes compiled on Kindle. The experiences are different, so they don't cannibalize each other

Posting Pace

  • 3-4 videos per week after the boost period ends

  • A 2-week cycle of 4 episodes x 7 videos

  • Ideally, fill it with scheduled posts for one month

  • Add to the release schedule in the order that auto-dubbing succeeds

👉 Ultimately, it will be 3-4 distributions per week. Considering the publishing pace on the Kindle side, this is it. Currently, publishing in each language is lagging behind. (A matter of motivation, haha)

Next Steps

  • Update all video descriptions to include source information

  • Publish auto-dubbed videos at 12:00 AM on 4/4

  • Changed posting schedule to a fixed time of 12:00 AM

  • Proceed with production of new Volume 10 content on a video-first basis

  • Source format standardization communicated to the production team

Advertisement