The world began from ash. — Visualizing the Azuchi-Momoyama to Edo periods using folding screens and AI

I am LaLaLa, currently in the midst of trial and error.
It took a whole day to complete.
I used a variety of content to produce it, but
I wonder if there is anyone who will actually watch it...

I will light the gold in Nobunaga's Azuchi Castle just once. I will make that gold explode across the entire screen in Hideyoshi's Golden Tea Room. And then, I will return the gold to ash in Ieyasu's Siege of Osaka. The title "Ash and Gold" was chosen because of this cycle of gold. Dawn, Rise, Gold, Shadow, Collapse, Echo—the six chapters form an arc from sunrise to sunset. (Embed the completed YouTube video here) This video is driven by a single policy: "Do not explain." Place only the events, and do not depict the inner thoughts. Entrust the emotions to the burning castle, the helmet fallen in the mud, and the empty seat.

1. Why three hours became one hour — The meaning of running the pipeline once

The reason the first one is heavy is not because the art is bad or the composition is poor.It is because the pipeline of the process has not yet run through to the end even once.

A single video is made through this flow.

Generate images → Animate with morphing → Synthesize narration → Create music → Bundle everything together in editing.

These five processes each use different tools and have different quirks. With the first one, you stop repeatedly at each step, wondering, "Is this right?" The prompt doesn't produce the image you imagined. The folds of the folding screen stand out when turned into a video. The audio is too bright. Each time, you research, try, and go back. This is the true nature of the three hours.

But once you run it through, the "way to get the right answer" for each process remains in your hands. From the second one on, you just pour the material into the same pipeline. That is why it is fast.

There is only one lesson. You are not making a work with the first one; you are laying the pipeline.Do not be greedy for quality here; prioritize running it from end to end at least once. A rough video that you finished is lighter for everything that follows than a perfect one.

2. Fixed phrases to make AI produce the folding screen style — The full prompt and how to crush stumbling blocks

The art in this work is unified in a style that mixes folding screen paintings (gold leaf clouds, flat bird's-eye view, ink outlines) and ukiyo-e. I always attach a "fixed phrase" to the beginning of every cut to make the AI produce this consistently.

There was a big pitfall here.Different tools interpret the same words differently.

The fixed phrase I wrote initially included the specification "Japanese folding screen byōbu painting." When I passed this to one image AI, it produced a magnificent folding screen. However, when I passed the same specification to another AI, itwould draw "the object known as a folding screen."A folding screen placed in a tatami room, wooden frames, hinges, even the floor. This is a "photograph of a folding screen," not "the world inside a folding screen." When turned into a video, the frame appears and breaks the immersion.

The cause was the words "folding screen" / "byōbu." When these are present, the AI tries to draw "a piece of furniture with a frame."

The way to crush this is to replace the words. Here is the fixed phrase that actually worked.

Japanese gold-leaf screen-painting style, full-bleed edge to edge,
no frame, no border, no wooden panel, not a photograph of a screen,
a few sparse ink-wash cloud bands, minimal gold,
flat aerial perspective, no cast shadows,
ink outlines, muted indigo / sumi-ink / unbleached paper palette,
figures small as a crowd, faces not detailed,
sparse composition, no text, --ar 16:9

There are three points where it is effective.

First,deletebyōbu / folding screen. Cut off the root cause of objectification.

Second,addfull-bleed edge to edge and no frame, no border, no wooden panel. Command it to "fill the entire screen to the edges with the painting."

Third, I add a final push with "not a photograph of a screen." I explicitly forbid it by saying, "Do not make it a photo taken of a folding screen."

With these three points, the framing issue disappeared.

Another thing: image AI has a habit where "words placed earlier have a stronger effect." Therefore, I place the most important instructions at the front of the fixed phrase. If you move them to the back, they weaken.

3. The Blueprint of Gold ── Where to place gold and where to extinguish it across six chapters

The thing I spent the most time designing in this work was not the story or the visuals, but the placement of gold.

This is why the title is "Ash and Gold." I do not use gold merely as decoration to make the screen look luxurious. I make the gold itself a character that moves through time.

I designed the arc of gold across the six chapters as follows.

In Chapter 1: Dawn (Nobunaga), I conserve the gold. Because it is a folding screen style, gold clouds are unavoidable as background, but I still do not make gold the protagonist. Only in one cut of Azuchi Castle do I light up the "glow" of gold. Up to this point, gold is merely a foreshadowing.

In Chapter 2: Rise (Hideyoshi), I unleash the gold. Gold first "strikes" the stone walls of Osaka Castle, and then, in the Golden Tea Room, the entire screen becomes gold. A portable golden room where all three mats are covered in gold leaf. This is the brightest frame in all six chapters. The explosion here is effective because I held back in the first chapter.

In Chapter 3: Gold (Ieyasu), I extinguish the gold. The Siege of Osaka. Toyotomi's castle burns, and the Golden Tea Room turns to ash. Ieyasu returns the gold, which reached its peak in Chapter 2, back to ash here. From this point on, gold disappears from the work.

From Chapter 4 onwards (Shadow, Collapse, Reverberation), it is a cold-colored world without gold. The peace loosens from within, the Black Ships arrive, and Edo becomes ruins.

In other words, gold follows a life cycle of conservation → explosion → extinction → absence. This perfectly overlaps with the shift in power from Nobunaga to Hideyoshi to Tokugawa. If you follow the gold, you can tell who holds the world.

How to make it effective in actual work is simple. Only for the key shots (Azuchi Castle, Golden Tea Room, Osaka Castle burning), I replace the fixed phrase "minimal gold" with "maximal radiant gold" or "burning gold." For all other cuts, I suppress the gold. I create the contrast between holding back and exploding with a single word in the prompt.

One caution. Even with the same "gold," you must shoot the gold of splendor and the gold that burns and disappears as separate things. The gold of the Golden Tea Room is "luxurious," while the gold of the burning Osaka Castle is "destructive." If you shoot the latter luxuriously, the death of the gold will not be conveyed.

4. How to create a cold narration ── Writing style and synthetic voice settings

There is only one principle for narration: do not explain.

Present only the events and do not describe the inner state. Entrust emotions to "objects" such as a burning castle, a helmet fallen in the mud, or an empty seat. Do not write "sad." Show the objects that harbor sadness.

For example, I wrote Honno-ji like this.

I do not write "Nobunaga died." I only place the fact that "the head was not found." The reader perceives the death from that absence.

There are three rules for writing.

First, keep sentences short. Long explanatory text kills the cold atmosphere.

Second, place words that act as sound effects on their own separate lines. "--Sudden rain." "--Gunshot." Do not have these read aloud; synchronize them with sound effects instead. Having them read aloud kills the tension.

Third, do not fear empty lines. Pauses (silence) are moments where unwritten emotions flow.

Settings for synthesized voice (VOICEVOX) are also more effective than the choice of voice itself.

Choose a voice quality that is low and crawling. I used Aoyama Ryusei's "sadness" style. The key is to reduce intonation. If left at the default, it expresses too much emotion. By lowering the intonation to less than half (around 0.4), the acting disappears, resulting in a "frozen voice." This brings it closer to the emotion-rejecting voice of a documentary.

As a guideline for settings, use an intonation of 0.4, a speaking speed of 0.85, and keep the pitch as is. It is more flexible to not linger on pauses in the voice software, but rather to extend them in editing software after exporting.

Since proper nouns are often misread, fix the pronunciation in advance. This alone will reduce accidents.

This concludes the design and process for the three volumes of "Ash and Gold."

One last thing. The true value of this production method is not that you can "create historical footage." It is deciding on one style and one principle, and sticking to them throughout the entire process. The arc of gold. Prose that does not explain. Moving folding screens. As long as these three pillars are established, you can build the same house even if you change the era.

From the fourth chapter onwards, you can stack them mechanically using the same design. That is what I do.

CRIMSON LOGIC / LaLaLa Books "Ash and Gold" -- Ash and Gold

.