An AI skill that turns books into animated, narrated, interactive web experiences
JavaScript ai-agentsai-educationanimationclaudeclaude-codeclaude-skillsedtecheducationinteractive-textbookscormtext-to-video
docs(site): the making-of pages of LRRH, the Joke Collection and the Pigs, cut to about half
yzr1cdc2f3b767e
diff --git a/site/joke-collection/index.html b/site/joke-collection/index.htmlindex c38d95d..f431187 100644--- a/site/joke-collection/index.html+++ b/site/joke-collection/index.html@@ -105,41 +105,32 @@</div><h2>Two ways, then a third</h2>- <p><a href="../three-little-pigs/">The Three Little Pigs</a> was painted by an image model, characters included. The world looked lovely, but every new action meant another round of drawings, and keeping a character the same from shot to shot was a fight. <a href="../little-red-riding-hood/">Little Red Riding Hood</a> went the other way: everything was paper drawn in code. Any movement was a few lines, but the world looked plain.</p>- <p>So we split the work by one question: <b>does it move?</b> Anything that stays put (streets, walls, blackboards, desks, the bun cart) is painted by the image model. Anything with joints, a face or a change of shape (the people, the falling rain, a bitten bun, the hands of the clock) is a paper puppet moved by code. A painting never has to act, and a puppet never has to look like a painting.</p>+ <p><a href="../three-little-pigs/">The Three Little Pigs</a> was painted by an image model, characters included: the world looked lovely, but every new action meant another round of drawings. <a href="../little-red-riding-hood/">Little Red Riding Hood</a> was all paper drawn in code: any movement was a few lines, but the world looked plain.</p>+ <p>So we split the work by one question: <b>does it move?</b> Anything that stays put (streets, walls, desks, the bun cart) is painted by the image model. Anything with joints, a face or a change of shape (the people, the rain, a bitten bun, the hands of the clock) is a paper puppet moved by code. A painting never has to act, and a puppet never has to look like a painting.</p><figure><img src="making/layers.webp" width="1400" height="879" alt="One frame taken apart: the painted street, the painted cart cut out on its own, the puppets drawn in code, and the finished frame." loading="lazy"><figcaption>One frame, taken apart. The cart is painted as a separate cut-out, so the bun seller can stand behind it.</figcaption></figure><h2>Finding the look</h2>- <p>Our puppets are flat paper with round heads, in the spirit of South Park. We tried three painted styles behind them: flat paper like the puppets (closest to that show), a warm gouache, and a richer torn-paper collage. We picked the collage. There was no reason to copy South Park exactly, and the point was to let the image model give the world as much texture and light as it could.</p>+ <p>Our puppets are flat paper with round heads, in the spirit of South Park. Behind them we tried flat paper, gouache and a torn-paper collage, and picked the collage, to let the image model give the world as much texture and light as it could.</p><figure><img src="making/styles-stall.webp" width="1400" height="788" alt="The bun stall four ways: drawn in code, flat paper, gouache, and torn-paper collage." loading="lazy"><figcaption>Top left: the stall drawn in code. Then flat paper (A), gouache (B), and torn-paper collage (C), the one we chose. Each test picture cost under two cents.</figcaption></figure>- <p>Every set was painted by redrawing our own code-drawn version in that style. That way the composition stays the same, and the puppets still stand where the scenes expect them. Signs and blackboards were painted blank, and their words were written on afterwards: image models still garble Chinese characters.</p>- <figure><img src="making/batch-stall.webp" width="1400" height="551" alt="The stall street in sun, rain and snow, and the painted props: buns, soy milk, umbrellas, homework sheets." loading="lazy"><figcaption>The stall in three weathers, from one picture, edited, so a cut between them changes only the weather. Below them, props painted on transparent backgrounds.</figcaption></figure>- <figure><img src="making/batch-sets.webp" width="1400" height="1050" alt="The town, the classroom from the front, the blackboard wall, the toilet, a pupil's desk and the teacher's desk." loading="lazy"><figcaption>The rest of the sets. The desks are separate pieces, because pupils sit behind them, row behind row. The clock is painted without hands, so code can turn them.</figcaption></figure>+ <p>Every set was painted by redrawing our own code-drawn version in that style, so the puppets still stand where the scenes expect them. Signs and blackboards were painted blank and written on afterwards: image models still garble Chinese characters.</p>+ <figure><img src="making/batch-stall.webp" width="1400" height="551" alt="The stall street in sun, rain and snow, and the painted props: buns, soy milk, umbrellas, homework sheets." loading="lazy"><figcaption>The stall in three weathers, edited from one picture, so a cut between them changes only the weather. Below, props on transparent backgrounds.</figcaption></figure>+ <figure><img src="making/batch-sets.webp" width="1400" height="1050" alt="The town, the classroom from the front, the blackboard wall, the toilet, a pupil's desk and the teacher's desk." loading="lazy"><figcaption>The rest of the sets. The desks are separate pieces, because pupils sit behind them. The clock is painted without hands, so code can turn them.</figcaption></figure><h2>What the code still does</h2><ul>- <li><b>The acting.</b> Mouths timed to every word, blinks at each puppet's own moments, eyes that follow whoever is talking, a whole class turning its heads a frame or two apart.</li>- <li><b>The joke's machinery.</b> Two thought bubbles (a bun and a toilet) pushing each other and merging, before Bo opens his mouth.</li>- <li><b>Weather and small things.</b> Rain and snow falling over the painted street, umbrellas, a fly, the clock's hands.</li>- <li><b>The film.</b> Cuts, close-ups, a soft background in close-ups, a soft shadow under every puppet so it stands in the painted light, and a gentle hand-made wobble.</li>+ <li><b>The acting:</b> mouths timed to every word, eyes that follow whoever is talking, a whole class turning its heads a frame or two apart.</li>+ <li><b>The joke's machinery:</b> two thought bubbles (a bun and a toilet) pushing each other and merging.</li>+ <li><b>Weather and small things:</b> rain and snow over the painted street, a fly, the clock's hands.</li>+ <li><b>The film:</b> cuts, close-ups, and a soft shadow under every puppet so it stands in the painted light.</li></ul><h2>A joke needs its setup</h2>- <p>The first version didn't land. Bo's slip ("Wanna go eat breakfast… in the bathroom?") came out of nowhere, and he looked as if he really meant it. The fix was story, not animation. First we show his two everyday questions: "Breakfast?" every morning at the stall, "Bathroom?" after every class. Then, the one morning he oversleeps and comes in starving, we show both wants at once, as two bubbles that collide. Only then does he speak. Same lesson as Little Red Riding Hood: set up everything the punchline needs, before anyone animates a frame.</p>+ <p>The first version didn't land. Bo's slip ("Wanna go eat breakfast… in the bathroom?") came out of nowhere. The fix was story, not animation: first we show his two everyday questions, "Breakfast?" every morning and "Bathroom?" after every class; then, the morning he comes in starving, both wants at once, as two bubbles that collide. Same lesson as Little Red Riding Hood: set up everything the punchline needs before anyone animates a frame.</p>- <h3>Episode 2, made in one night</h3>- <p>For the second joke, the user told the story at bedtime and handed over: a fresh session read this method, wrote the script, painted two sets (the store outside and in), a counter to stand behind and three props, built three new puppets and four scenes, and checked them, all before morning. The clerk's Japanese is spoken in Japanese with English subtitles, so the audience understands both sides while neither character does. Its painting cost $0.13.</p>-- <h3>Episode 3, made with the Skill</h3>- <p>The third joke was the test of the Papermorph Skill itself. A new session started from nothing but the Skill and the joke, asked a few questions (which language each man speaks, whether English viewers need "lying flat" explained), wrote the script, painted two country roads and a bundle of hay, built a cow and three new puppets, and checked its own work. Each man speaks his own language, Mandarin, Hindi and English, with English subtitles. Its painting cost $0.07.</p>-- <h3>Episode 4, filmed twice</h3>- <p>The fourth joke was cut right after the first one, then shelved: the story worked, the filming didn't. Every object the jokes hang on (an oil drop, a blank sheet, three bites) was a few pixels in a wide shot, and the main gag played in one locked frame. A review compared it shot by shot with the episodes above, and four agents filmed each scene again: the main scene went from 6 shots to 22, with a close-up for every object a joke needs and a reaction for every joke. Same script, same voices, no new paintings.</p>-- <h3>Episode 5, nobody in the loop</h3>- <p>For the fifth, the user said only: make one, without me. Claude picked an old joke, and a coordinator ran the rest with agents: one wrote the script and built the harbour, the street, the beach and three new puppets; four filmed a scene each; then a director, two picture reviewers and a sound reviewer watched every scene against the episodes above and sent them back, three rounds in all. The running gag (he stops, the penguin stops half a step late) came out of the first review. Its painting cost $0.18.</p>- <p>For the sixth, the user brought the joke Ronald Reagan used to tell about three dogs, and added a fourth. They changed the last line so it rhymes ("Shh. Don't bark, or you are the meat."), asked for it all in English, and after watching the first cut said the Polish dog shouldn't carry his bowl and the barks sounded odd, so the dogs now shout their own "Woof!". The miscount became the opening gag, and a cook's shadow chopping behind a kitchen door became the reason nobody barks. The same team of agents ran three review rounds. Its painting cost $0.14.</p>- <p>The seventh came from outside: <a href="https://github.com/yuyays">Yuya</a> made it with the same Skill and sent it in as a pull request. Spain’s everyday word for a cow, “vaca”, sounds to Yui like “baka”; three new painted sets (a farm gate, a pasture and a stable), and the letters on the stable board drawn in code over a blank painted board, so a typo never costs a repaint. Two more rounds of review fixes came in on the pull request before it went on the shelf.</p>+ <h3>The later episodes</h3>+ <p><b>Episode 2</b> was made in one night: the user told the story at bedtime, and a fresh session wrote, painted, built and checked it before morning. The clerk speaks Japanese with English subtitles, so the audience understands both sides while neither character does. <b>Episode 3</b> tested the Papermorph Skill itself: a new session started from nothing but the Skill and the joke, asked a few questions, and checked its own work.</p>+ <p><b>Episode 4</b> was filmed twice: the story worked, but every object the jokes hang on was a few pixels in a wide shot. Four agents filmed each scene again, and the main scene went from 6 shots to 22, with no new paintings. For <b>episode 5</b> the user said only: make one, without me. A coordinator ran agents, and a director and reviewers sent scenes back for three rounds; the running gag (he stops, the penguin stops half a step late) came out of the first review.</p>+ <p><b>Episode 6</b> is the joke Ronald Reagan used to tell about three dogs, with a fourth added and a last line that rhymes; after the first cut the user asked for the dogs to shout their own "Woof!". <b>Episode 7</b> came from outside: <a href="https://github.com/yuyays">Yuya</a> made it with the same Skill and sent it in as a pull request.</p><h2>The numbers</h2><div class="numbers">diff --git a/site/little-red-riding-hood/index.html b/site/little-red-riding-hood/index.htmlindex 1e82ebf..9dfe31c 100644--- a/site/little-red-riding-hood/index.html+++ b/site/little-red-riding-hood/index.html@@ -85,59 +85,50 @@</div><h2>What we set out to do</h2>- <p>The Three Little Pigs showed that an image model can paint a lovely world, but every new action needed a new round of drawings, and the characters had to be fought into consistency. This time we asked the opposite question: what if the characters stay simple shapes drawn in code, and we lean into that, so that they look like paper puppets on purpose?</p>- <p>After the wolf runs off, the film stops in the meadow for children learning English. Little Red Riding Hood asks for one thing at a time ("Can you find a flower for Grandma? Touch it!"), and every flower, tree, cloud and creature answers a tap with its name. Nothing is ever called wrong: each miss helps a little more, the thing she wants wiggles and paper stars circle it, and if nobody finds it she points it out herself. A flower you find flies into her bunch for Grandma.</p>- <p>The story is the Brothers Grimm version (1857), retold lightly: the wolf still talks his way into Grandma's house, but Grandma hides in the wardrobe, nobody gets eaten, and the wolf leaves through the window in Grandma's nightcap.</p>+ <p>The Three Little Pigs showed that an image model can paint a lovely world, but every new action needed new drawings. So we asked the opposite question: what if the characters stay simple shapes drawn in code, and look like paper puppets on purpose?</p>+ <p>The story is the Brothers Grimm version (1857), retold lightly: Grandma hides in the wardrobe, nobody gets eaten, and the wolf leaves through the window in Grandma's nightcap. In the meadow the film stops for children learning English: Little Red Riding Hood asks for one thing at a time ("Can you find a flower for Grandma? Touch it!"), everything answers a tap with its name, and each miss helps a little more.</p><h2>Making it look like stop motion</h2><ul>- <li><b>Twelve frames a second, on twos.</b> Every movement is sampled into steps, so it moves in little jumps, the way a hand-moved puppet does.</li>- <li><b>The animator's hand.</b> Each frame nudges every puppet by a tiny random amount, and the picture flickers slightly. The nudge belongs to the frame, so a paused or rewound film looks exactly the same.</li>- <li><b>Paper you can feel.</b> Each piece has paper grain, torn edges and a soft shadow, and the joints show brass pins.</li>- <li><b>Replacement, not morphing.</b> Mouths are swapped like stop-motion mouths, blinks are two closed-eye frames, and a door opens through a few boards of different widths.</li>+ <li><b>Twelve frames a second, on twos,</b> so it moves in little jumps, the way a hand-moved puppet does.</li>+ <li><b>The animator's hand:</b> each frame nudges every puppet by a tiny random amount.</li>+ <li><b>Paper you can feel:</b> grain, torn edges, a soft shadow, brass pins at the joints.</li>+ <li><b>Replacement, not morphing:</b> mouths are swapped like stop-motion mouths.</li></ul><h2>The cast</h2>- <p>Five puppets: Little Red Riding Hood, her mother, the wolf (flat cap, red bow tie, more polite than he should be), Grandma (nightcap, glasses, umbrella) and the woodcutter (check shirt, axe). Each puppet is a handful of paper pieces pinned at the joints. A pose is just a set of joint angles, so a new action, like ducking, swinging an axe or diving through a window, costs a few numbers instead of new drawings.</p>+ <p>Five puppets, including a wolf in a flat cap and red bow tie, more polite than he should be. A pose is just a set of joint angles, so a new action, like ducking or diving through a window, costs a few numbers instead of new drawings.</p><figure><img src="making/cast.webp" width="1400" height="455" alt="The five puppets, each seen from the side and from the front." loading="lazy"><figcaption>Every puppet comes as two complete sets: side view and front view.</figcaption></figure>- <h3>Turning around is a whole-body swap</h3>- <p>A paper puppet can't rotate, so a turn works like a cut-out cartoon: the side view squashes, the whole front view takes over for the narrowest frames, and the mirrored side view comes out the other way. Head and body always turn together. A front head on a side body looked wrong every time.</p>+ <p>A paper puppet can't rotate, so a turn swaps whole views: side, front, mirrored side. Head and body always turn together; a front head on a side body looked wrong every time.</p><figure><img src="making/turn.webp" width="1400" height="420" alt="Little Red Riding Hood turning: side view squashing, front view, mirrored side view." loading="lazy"></figure>- <h3>Mouths, walks and hands</h3>- <p>Nine mouth shapes, each timed to the words of the voice track. Walks use a compass step: the planted foot stays put while the hip swings over it, so feet never slide on the ground. When someone hands something over, both hands really touch it.</p>+ <p>Nine mouth shapes are timed to the voice track, and walks use a compass step, so feet never slide.</p><figure><img src="making/mouths.webp" width="1400" height="245" alt="Little Red Riding Hood's nine mouth shapes." loading="lazy"><figcaption>The nine mouths, from rest to "oo".</figcaption></figure><figure><img src="making/walk.webp" width="1400" height="350" alt="Six frames of the wolf's walk." loading="lazy"><figcaption>Six frames of the wolf's walk.</figcaption></figure>- <h2>Sets and the camera</h2>- <p>Nine sets, each drawn as layers of paper: sky, distant hills, trees, the building, and plants close to the lens. As the camera moves, the layers slide past each other at different speeds, and in close-ups the background goes soft. Windows and doorways are real holes, so we can see people inside a house from outside. Reverse shots get their own sets, so the background behind a face is always what that person is looking at.</p>- <figure><img src="making/sets.webp" width="1400" height="395" alt="Eight sets: Mother's cottage, the garden, two woods, Grandma's cottage, her bedroom and two reverse angles." loading="lazy"><figcaption>Mother's cottage and garden, the woods both ways, Grandma's cottage, her bedroom, and the bedroom seen from the bed and from the door.</figcaption></figure>+ <h2>Sets, camera and sound</h2>+ <p>Each set is layers of paper that slide past each other as the camera moves. Windows and doorways are real holes, so we can see inside a house from outside.</p>+ <figure><img src="making/sets.webp" width="1400" height="395" alt="Eight sets: Mother's cottage, the garden, two woods, Grandma's cottage, her bedroom and two reverse angles." loading="lazy"><figcaption>The eight sets, two of them reverse angles of Grandma's bedroom.</figcaption></figure><figure><img src="making/layers.webp" width="1400" height="172" alt="Grandma's cottage taken apart into sky, far, mid, inside and house layers." loading="lazy"><figcaption>Grandma's cottage, layer by layer. "Inside" is the room seen through the door and window holes.</figcaption></figure>- <p>Shots change as in a film: wide shots to show where we are, two-shots for conversations, a close-up for whoever speaks, and each question to "Grandma" closer than the last. At a few scene changes, a big tree trunk sweeps across the frame, as in a paper theatre.</p>-- <h2>Voices, music and sound</h2>- <p>Six voices (the narrator, Little Red Riding Hood, Mother, the wolf, Grandma and the woodcutter) are ElevenLabs voices that act: the wolf squeaks like a little girl at Grandma's door, quavers like an old lady in bed, and drops into his own roar on "EAT". The words in each line are timed by listening to it, and those times drive the mouths and the captions. The music is the Three Little Pigs' score, reused. Every sound effect is synthesized in code: beds for the woods, the meadow and the bedroom, footsteps and paws, the knock, the latch and the creaking door, the axe (muffled when we are indoors), the wardrobe slam, the sting before "EAT", the umbrella on the wolf's head and the window breaking.</p>+ <p>Shots change as in a film, each question to "Grandma" closer than the last. The six ElevenLabs voices act: the wolf squeaks like a little girl at Grandma's door and drops into his own roar on "EAT". The music is the Three Little Pigs' score, reused, and every sound effect is synthesized in code.</p><h2>The rescue that didn't make sense</h2>- <p>In the first cut, the woodcutter appeared from nowhere, Grandma suddenly stood outside the wardrobe, and the wolf flew out of the window as if Grandma had superpowers. Every shot looked fine. The story didn't hold together. We rewrote it so that each moment causes the next:</p>- <figure><img src="making/chop.webp" width="800" height="450" alt="Little Red Riding Hood greets the woodcutter chopping wood near Grandma's cottage." loading="lazy"><figcaption>The woodcutter now appears earlier. He is chopping wood near the cottage, and Little Red Riding Hood says good morning.</figcaption></figure>+ <p>In the first cut, the woodcutter appeared from nowhere and the wolf flew out of the window as if Grandma had superpowers. Every shot looked fine; the story didn't hold together. We rewrote it so that each moment causes the next:</p>+ <figure><img src="making/chop.webp" width="800" height="450" alt="Little Red Riding Hood greets the woodcutter chopping wood near Grandma's cottage." loading="lazy"><figcaption>The woodcutter now appears earlier, chopping wood near the cottage.</figcaption></figure><figure><img src="making/rescue.webp" width="1400" height="526" alt="Six moments of the rescue." loading="lazy"></figure><ol class="beats">- <li>The wolf leaps; she ducks and cries "Help!"; he snaps up her flowers and lands face first.</li>- <li>Earlier, while he said "all the better to see you with", we glimpsed Grandma's eye in the crack of the wardrobe door. Now the door bursts open.</li>- <li>She runs at him and brings her umbrella down on his head.</li>+ <li>The wolf leaps; she ducks and cries "Help!"</li>+ <li>Grandma bursts out of the wardrobe and brings her umbrella down on his head.</li><li>He bolts for the front door, where the woodcutter, who heard the cry, steps in.</li>- <li>Grandma on one side, the woodcutter at the door: only one way out.</li>- <li>Through the window. Outside, the window stays broken.</li>+ <li>Only one way out: through the window.</li></ol>- <p>The lesson: check the script for cause and effect before animating. A story fix is cheap on paper, and expensive once every shot is built.</p>+ <p>A story fix is cheap on paper, and expensive once every shot is built.</p><h2>Everything we made for it</h2><dl class="made">- <dt>Puppets</dt><dd>Five characters, side and front sets each, with eyes, brows, nine mouths and fingers that stay live while the paper pieces are baked.</dd><dt>Sets</dt><dd>Eight layered sets, each with a blurred copy for close-ups.</dd>- <dt>Props</dt><dd>Basket with cake and apple juice, a bouquet that grows as she picks it, Grandma's nightcap and glasses, umbrella, axe and chopping block, quilt, doors that swing, window glass that shatters into falling shards, the spilled flowers.</dd>- <dt>Sound</dt><dd>Six acted voices, nine mouth shapes per line, four music cues, three synthesized beds and about twenty effects.</dd>- <dt>Tools</dt><dd>A set drawer and baker; a puppet baker that turns all 230 paper pieces into images so the film starts at once; a sound synthesizer; voice, mouth and audio-joining scripts reused from the pigs; a cast lab for posing puppets; and contact sheets, a few frames of every shot side by side, checked before anyone else watched.</dd>+ <dt>Props</dt><dd>Basket, a bouquet that grows as she picks it, umbrella, axe, doors that swing, window glass that shatters.</dd>+ <dt>Sound</dt><dd>Six acted voices, four music cues, three synthesized beds and about twenty effects.</dd>+ <dt>Tools</dt><dd>A puppet baker that turns all 230 paper pieces into images so the film starts at once, a sound synthesizer, and contact sheets of every shot, checked before anyone else watched.</dd></dl><h2>The numbers</h2>@@ -150,22 +141,15 @@</div><h2>The second version</h2>- <p>Twelve days later, <a href="../little-horse/">The Little Horse Crosses the River</a> set a higher bar, and we brought this film up to it. The script, the cast and the paper look stayed; the voices were redone with actors' delivery, and every scene was restaged by a team of agents, each taking one scene, with a director reviewing the whole film twice:</p>- <ul>- <li><b>The turning point is seen.</b> The wolf points at a real meadow; she looks at the path, a picture of Mother's raised finger flickers past, and then we see her foot step off the path. Before, she was sixty pixels high and never left it.</li>- <li><b>Every story point gets its shot.</b> The basket with the cake and juice, the three oaks (now three whole trees you can count), Grandma's cap and glasses left on the pillow, the wide-open door, the ears, the eyes and the teeth.</li>- <li><b>Grandma's escape reads.</b> We follow her hop into the wardrobe, and the wolf has his back to it, so a child can see why he never opens it.</li>- <li><b>The questions build.</b> She gets more worried each time, the music stops, and only then: "EAT".</li>- <li><b>The meadow stop became a game.</b> See above: the character asks, the picture answers.</li>- </ul>+ <p>Twelve days later, <a href="../little-horse/">The Little Horse Crosses the River</a> set a higher bar, and we brought this film up to it. The voices were redone, and a team of agents restaged every scene, with a director reviewing the whole film twice. Now we see her foot step off the path, every story point gets its shot, and the questions build until the music stops, and only then: "EAT".</p><h2>Where this leaves us</h2><div class="verdict">- <p><b>New actions are cheap.</b> The pigs could only do what had been drawn. Here, the umbrella blow, the duck and the dive through the window were each a few lines, added on the last evening.</p>- <p><b>Consistency comes free.</b> A puppet is the same paper in every shot, so nothing drifts between scenes.</p>+ <p><b>New actions are cheap.</b> The umbrella blow and the dive through the window were each a few lines.</p>+ <p><b>Consistency comes free.</b> A puppet is the same paper in every shot.</p><p><b>It gets cheaper with practice.</b> Puppets, sets, sounds and tools carry over to the next story.</p></div>- <p>The limits are real too. The look is deliberately simple, the paper effect has to be baked ahead of time to play smoothly, and with no painter to hide behind, the story itself has to hold up. We plan to build the Papermorph Skill's picture-book method on this.</p>+ <p>The limits are real too: the look is deliberately simple, and with no painter to hide behind, the story itself has to hold up. We plan to build the Papermorph Skill's picture-book method on this.</p><p><a class="watch" href="ch01/">▶ Watch the film</a></p></main><footer>Story: the Brothers Grimm's <i>Little Red Cap</i> (1857), retold for this film. No image model: everything you see is drawn in code. Music from Google Lyria (made for The Three Little Pigs), voices by ElevenLabs (v4), sound effects synthesized. Part of <a href="../">Papermorph</a>.</footer>diff --git a/site/three-little-pigs/index.html b/site/three-little-pigs/index.htmlindex 9c53308..4066523 100644--- a/site/three-little-pigs/index.html+++ b/site/three-little-pigs/index.html@@ -85,15 +85,15 @@</div><h2>What we set out to do</h2>- <p>Take a story with no book and no pictures, and make it an animated film you can pause, scrub and read along with. The pages are the same kind as the rest of the shelf: static web pages, one SVG stage, narration clips driving every move. What was new: the art came from an image model (OpenAI's gpt-image through OpenRouter), and the goal was for it to feel like a TV cartoon, not a slideshow.</p>+ <p>Take a story with no book and no pictures, and make it an animated film you can pause, scrub and read along with, built like the rest of the shelf. What was new: the art came from an image model (OpenAI's gpt-image through OpenRouter), and the goal was a TV cartoon, not a slideshow.</p><figure><img src="making/styles.webp" width="1400" height="260" alt="The same scene in three styles: gouache, cut paper, TV cartoon." loading="lazy"><figcaption>Style tests, one scene in three looks. We chose the TV cartoon (right).</figcaption></figure><h2>Four tries at a character</h2><ol class="tries">- <li><b>Rotating painted parts.</b> We cut the wolf into image pieces and rotated them at the joints. Too stiff: it read as a paper puppet, with seams at every knee.</li>- <li><b>Painted body, code-drawn limbs.</b> Arms and legs became bendable SVG curves with feet planted by inverse kinematics. Much livelier, but the first scene, shot from one fixed camera, still felt like a slideshow.</li>- <li><b>Separately painted parts.</b> Heads, bodies, sleeves and hands from separate generations never lined up: arms came out beside the sleeve, collars doubled, hands went missing.</li>- <li><b>Characters drawn entirely in code.</b> They lined up perfectly, but looked far too crude next to the painted world.</li>+ <li><b>Rotating painted parts.</b> Too stiff, with seams at every knee.</li>+ <li><b>Painted body, code-drawn limbs.</b> Much livelier, but one fixed camera still felt like a slideshow.</li>+ <li><b>Separately painted parts.</b> They never lined up: arms came out beside the sleeve, hands went missing.</li>+ <li><b>Characters drawn entirely in code.</b> Far too crude next to the painted world.</li></ol><div class="pair"><figure><img src="making/try1-rigid.webp" width="1400" height="788" alt="The wolf assembled from rigid painted parts." loading="lazy"><figcaption>Try 1: rigid parts.</figcaption></figure>@@ -104,32 +104,31 @@<h2>What finally worked</h2><h3>One picture, edited, never redrawn</h3>- <p>Each character got one large picture. Every other version was an <em>edit</em> of that same picture: without the near arm, without eyes and mouth, with the legs in each walking pose, with each mouth shape. Because the edits change only a few percent of the pixels, they line up exactly, and a script can cut parts out wherever two pictures differ.</p>+ <p>Each character got one large picture, and every other version was an <em>edit</em> of it: without the near arm, without a face, with the legs in each walking pose. The edits change only a few percent of the pixels, so they line up exactly, and a script can cut parts out wherever two pictures differ.</p><figure><img src="making/edits-mother.webp" width="1400" height="689" alt="Mother Pig three times: complete, without her arm, without her face." loading="lazy"><figcaption>Mother Pig as drawn, without her arm (the dress behind it painted in), and without her face.</figcaption></figure><h3>Vectorized, then swapped like a TV cartoon</h3>- <p>The cut parts were traced into vector shapes, so close-ups stay sharp. Then they were baked back to images, because live tracing was too slow. Movement comes from swapping drawings the way TV cartoons do. Legs walk on four drawings, each held for a few frames. Arms swap between drawn gestures. Big actions, like sleeping or running, are whole drawings of their own. Code adds the rest: body bounce, blinks, eye movement and the camera.</p>+ <p>The cut parts were traced into vector shapes, so close-ups stay sharp. Movement comes from swapping drawings the way TV cartoons do: legs walk on four drawings, and big actions, like sleeping or running, are whole drawings of their own. Code adds body bounce, blinks and the camera.</p><figure><img src="making/vectorize.webp" width="1400" height="741" alt="Mother Pig painted, vectorized, and zoomed in three times." loading="lazy"><figcaption>Painted, vectorized, and the vector zoomed in three times.</figcaption></figure><figure><img src="making/walk-wolf.webp" width="1400" height="516" alt="Four walking drawings of the wolf." loading="lazy"><figcaption>The wolf's walk: four drawings, all edits of the same picture.</figcaption></figure><figure><img src="making/whole-poses.webp" width="1400" height="410" alt="The first pig carrying, tossing, napping, laughing and running." loading="lazy"><figcaption>Whole-body drawings for actions that limbs alone can't do.</figcaption></figure><figure><img src="making/wolf-poses.webp" width="1400" height="292" alt="The wolf sniffing, grinning, inhaling, blowing, tiptoeing, collapsing and howling." loading="lazy"><figcaption>Every wolf action in the film is one of these.</figcaption></figure><h3>Faces that talk</h3>- <p>Each face has a featureless "faceplate" laid over it. On top go moving eyes, lids that slide down to blink, and brows. Mouth shapes are painted by the model, and each one is timed to a word from the voice track.</p>+ <p>Each face has a featureless "faceplate" laid over it, with moving eyes, lids and brows on top. Painted mouth shapes are timed to the words of the voice track.</p><figure><img src="making/mouths-wolf.webp" width="1400" height="245" alt="The wolf's mouth: closed, and four open shapes." loading="lazy"><figcaption>The wolf's mouths. Each is a block of the muzzle with a soft edge, so the jaw can really open.</figcaption></figure><figure><img src="making/mouths-mother.webp" width="1400" height="195" alt="Mother Pig's mouth: closed and four shapes." loading="lazy"></figure><h2>Places, houses and the camera</h2>- <p>Backgrounds were painted, then separated into layers: sky, hills, ground, buildings, and flowers close to the camera. Layers slide past each other as the camera moves, and the background softens in close-ups. The houses come in three aligned stages, so building them is a series of thumps, not a wipe. The camera cuts on words: a close-up for whoever speaks, a follow shot when someone runs. Silences were spliced into the narration to give the building and the huffing room to happen.</p>+ <p>Backgrounds were painted, then separated into layers that slide past each other as the camera moves. The houses come in three aligned stages, so building them is a series of thumps, not a wipe. The camera cuts on words: a close-up for whoever speaks, a follow shot when someone runs.</p><figure><img src="making/layers.webp" width="1600" height="178" alt="The cottage scene as five separate layers." loading="lazy"><figcaption>Mother Pig's cottage in five layers.</figcaption></figure><figure><img src="making/houses.webp" width="1600" height="175" alt="Each house in three construction stages." loading="lazy"><figcaption>Three houses, three stages each.</figcaption></figure>- <p>The voices are free text-to-speech, one voice per character; when all three pigs answer the wolf, three voices are stretched to the same length and mixed into a chorus. The music, four instrumental pieces, was generated with Google's Lyria and crossfades between parts of the story, dipping under the voices.</p>+ <p>The voices are free text-to-speech, one per character. The music, four pieces generated with Google's Lyria, crossfades between parts of the story.</p><h2>What kept going wrong</h2>- <p>Almost every fix came from a problem spotted while watching, and many of them were specific to one character.</p>+ <p>Almost every fix came from watching, and many were specific to one character.</p><figure><img src="making/bugs-hands.webp" width="1400" height="344" alt="A hollow arm, a missing hand, and ghost legs." loading="lazy"><figcaption>A see-through arm (fur the same colour as the body behind it fooled the cutting), a missing hand, ghost legs from blending two drawings.</figcaption></figure><ul>- <li>The second pig wore no trousers in his first design. The image model's safety filter kept refusing his pictures. Another model redrew each one slightly differently, so his body flickered as he walked. In the end he got shorts and was redrawn from scratch.</li>- <li>A held object belongs to one drawing: the third pig carried the pot while still holding a brick, because his carrying arm was the brick-carrying drawing.</li>- <li>Anything without its own drawing slides instead of moving, like the wolf running away on a single pose, until we gave it the walking legs.</li>+ <li>The second pig wore no trousers, and the image model's safety filter kept refusing his pictures. Another model redrew each one slightly differently, so his body flickered. In the end he got shorts and was redrawn from scratch.</li>+ <li>A held object belongs to one drawing: the third pig carried the pot while still holding a brick.</li><li>Quality "high" cost five times "medium" and, once traced, looked the same. We found out late.</li></ul><figure><img src="making/medium-vs-high.webp" width="1400" height="788" alt="Medium and high quality side by side, zoomed in." loading="lazy"><figcaption>Medium (left) and high (right), vectorized and zoomed in three times.</figcaption></figure>@@ -144,11 +143,11 @@<h2>Why we don't recommend it</h2><div class="verdict">- <p><b>Every new action needs new drawings.</b> In a book drawn in code, a new movement is a few lines. Here, picking up a pot, climbing a ladder or running away each means another round of edits per character, plus cutting, checking and fixing. The film can only do what was drawn, and you notice the gaps.</p>- <p><b>Consistency has to be fought for, character by character.</b> Hips, collars, sleeves, fur and mouths each needed their own handling. A safety filter or a second model could break a character halfway through.</p>- <p><b>It doesn't get cheaper with practice.</b> The first scenes were polished. The later ones, made faster, came out rough. A Skill should make the hundredth chapter as good as the first. This method makes each one cost about the same, in money and attention.</p>+ <p><b>Every new action needs new drawings.</b> In a book drawn in code, a new movement is a few lines. Here, picking up a pot or climbing a ladder means another round of edits per character. The film can only do what was drawn, and you notice the gaps.</p>+ <p><b>Consistency has to be fought for, character by character.</b> A safety filter or a second model could break a character halfway through.</p>+ <p><b>It doesn't get cheaper with practice.</b> A Skill should make the hundredth chapter as good as the first. This method makes each one cost about the same, in money and attention.</p></div>- <p>So the story is finished and stays here, as a record of a path we walked and wouldn't recommend as it stands. What we kept from it: an image model is very good at painting a world that stays still. <a href="../joke-collection/">Joke Collection</a> uses it for exactly that, sets and props only, while paper puppets drawn in code do all the acting. What would change our mind: image models that can hand back a character as a real, bendable rig, or video models that keep a character consistent and can be steered word by word.</p>+ <p>So the film stays here as a record of a path we wouldn't recommend as it stands. What we kept: an image model is very good at painting a world that stays still, and <a href="../joke-collection/">Joke Collection</a> uses it for exactly that. What would change our mind: image models that hand back a character as a real, bendable rig, or video models that keep a character consistent and can be steered word by word.</p><p><a class="watch" href="ch01/">▶ Watch the film</a></p></main><footer>Story: the English folk tale, retold for this film (in this version nobody gets eaten). Pictures generated with OpenAI gpt-image and Google Gemini image models, music with Google Lyria, voices with Microsoft Edge text-to-speech. Part of <a href="../">Papermorph</a>.</footer>