Character serial, block 2
The same two characters on a different frame: the flyer stays upside down from the lamp cord through three lines while a stool is set down.






4 models received different prompt text — show each
Each model is given the prompt its input contract calls for: a multi-reference model is told which reference is which character, a segmented model gets shot segments. The frame, the references and the requested length are identical throughout. This is the honest version of "the same prompt".
H3 Max, Wan 2.7, Veo 3.1, LTX-2
Modern 3D animated feature style for preschool: soft rounded forms, chunky simple silhouettes, large expressive eyes, smooth matte surfaces with subtle fuzz on fur, warm rim light and soft ambient occlusion, shallow depth of field, saturated but gentle colour, no harsh shadows, no photorealism. Cosy and glowing rather than dark. Vertical 9:16, character centred and large in frame, uncluttered background so the character always reads instantly. A single continuous 10-second scene. One location: a tiny warm bakery at night. A wooden counter, a glass dome over a cake stand, a round brass bell, shelves of jars, a big friendly oven with a glowing window, and one window onto a deep blue night with round stars. Everything is soft and rounded - no sharp knives, no fire, no mess. Characters are stylized animals that walk upright, wearing at most one simple accessory each. No humans, no text or signage, no crowds. The night outside is cosy, never frightening. 0s–5s: The baker sets a little stool down and pats it. The flyer, still upside down: 'That's very kind. But then I'd be the wrong way up.' The baker nods slowly and puts the stool away. 5s–10s: The baker hums and taps the counter twice with one claw. 'The jam is the problem.' The flyer, delighted: 'That is what I keep saying.' The baker's eyebrows go up. He turns and gets a bun. Hold every character's design exactly as shown. Do not drift toward photorealism. Avoid: photorealistic, live action, human, person, realistic fur, uncanny, creepy, scary, dark, gloomy, horror, gritty, harsh shadow, sharp knife, fire, flames, mess, spilled, sharp teeth, text, letters, signage, watermark, subtitles, caption, logo, cluttered background, crowd, extra characters, deformed, extra limbs, distorted face
H3 ref2v, Seedance 2.0
Modern 3D animated feature style for preschool: soft rounded forms, chunky simple silhouettes, large expressive eyes, smooth matte surfaces with subtle fuzz on fur, warm rim light and soft ambient occlusion, shallow depth of field, saturated but gentle colour, no harsh shadows, no photorealism. Cosy and glowing rather than dark. Vertical 9:16, character centred and large in frame, uncluttered background so the character always reads instantly. Image 1, Image 2, Image 3 are The baker (large brown bear baker, cream apron, round spectacles) from 3 angles — the same character every time. Image 4, Image 5, Image 6 are The flyer (small purple fruit bat, orange bow tie) from 3 angles — the same character every time. Image 7 is the room, and the exact look of the scene. A single continuous 10-second scene. One location: a tiny warm bakery at night. A wooden counter, a glass dome over a cake stand, a round brass bell, shelves of jars, a big friendly oven with a glowing window, and one window onto a deep blue night with round stars. Everything is soft and rounded - no sharp knives, no fire, no mess. Characters are stylized animals that walk upright, wearing at most one simple accessory each. No humans, no text or signage, no crowds. The night outside is cosy, never frightening. 0s–5s: The baker sets a little stool down and pats it. The flyer, still upside down: 'That's very kind. But then I'd be the wrong way up.' The baker nods slowly and puts the stool away. 5s–10s: The baker hums and taps the counter twice with one claw. 'The jam is the problem.' The flyer, delighted: 'That is what I keep saying.' The baker's eyebrows go up. He turns and gets a bun. Hold every character's design exactly as shown. Do not drift toward photorealism. Avoid: photorealistic, live action, human, person, realistic fur, uncanny, creepy, scary, dark, gloomy, horror, gritty, harsh shadow, sharp knife, fire, flames, mess, spilled, sharp teeth, text, letters, signage, watermark, subtitles, caption, logo, cluttered background, crowd, extra characters, deformed, extra limbs, distorted face
Kling v3 Pro
[{"prompt":"Baba sets a little stool down and pats it. Fen, still upside down: 'That's very kind. But then I'd be the wrong way up.' Baba nods slowly and puts the stool away.","duration":"5"},{"prompt":"Baba hums and taps the counter twice with one claw. 'The jam is the problem.' Fen, delighted: 'That is what I keep saying.' Baba's eyebrows go up. He turns and gets a bun.","duration":"5"}]H3 ref2v + voice, Seedance + voice
Modern 3D animated feature style for preschool: soft rounded forms, chunky simple silhouettes, large expressive eyes, smooth matte surfaces with subtle fuzz on fur, warm rim light and soft ambient occlusion, shallow depth of field, saturated but gentle colour, no harsh shadows, no photorealism. Cosy and glowing rather than dark. Vertical 9:16, character centred and large in frame, uncluttered background so the character always reads instantly. Image 1, Image 2, Image 3 are The baker (large brown bear baker, cream apron, round spectacles) from 3 angles — the same character every time. Image 4, Image 5, Image 6 are The flyer (small purple fruit bat, orange bow tie) from 3 angles — the same character every time. Image 7 is the room, and the exact look of the scene. Audio 1 is the scene's complete dialogue, in order: The flyer says "That's very kind. But then I'd be the wrong way up.", then The baker says "The jam is the problem.", then The flyer says "That is what I keep saying.". Animate each character's mouth to match Audio 1 exactly, and let the other characters listen while one speaks. A single continuous 10-second scene. One location: a tiny warm bakery at night. A wooden counter, a glass dome over a cake stand, a round brass bell, shelves of jars, a big friendly oven with a glowing window, and one window onto a deep blue night with round stars. Everything is soft and rounded - no sharp knives, no fire, no mess. Characters are stylized animals that walk upright, wearing at most one simple accessory each. No humans, no text or signage, no crowds. The night outside is cosy, never frightening. 0s–5s: The baker sets a little stool down and pats it. The flyer, still upside down: 'That's very kind. But then I'd be the wrong way up.' The baker nods slowly and puts the stool away. 5s–10s: The baker hums and taps the counter twice with one claw. 'The jam is the problem.' The flyer, delighted: 'That is what I keep saying.' The baker's eyebrows go up. He turns and gets a bun. Hold every character's design exactly as shown. Do not drift toward photorealism. Avoid: photorealistic, live action, human, person, realistic fur, uncanny, creepy, scary, dark, gloomy, horror, gritty, harsh shadow, sharp knife, fire, flames, mess, spilled, sharp teeth, text, letters, signage, watermark, subtitles, caption, logo, cluttered background, crowd, extra characters, deformed, extra limbs, distorted face
- The flyer: “That's very kind. But then I'd be the wrong way up.”
- The baker: “The jam is the problem.”
- The flyer: “That is what I keep saying.”
rendered in a production pipeline with a character bible and recorded cast voices — not reproducible from this repository. Character names and one spoken line were changed for publication, so takes rendered before that change say the original words and the lines check reads one short for them. Each take page shows the exact text its model received.
Same character, separate call: Character serial, block 1 · how consistent each model was
| H3 Max | H3 ref2v | Kling v3 Pro | Wan 2.7 | Seedance 2.0 | Veo 3.1 | LTX-2 | H3 ref2v + voice | Seedance + voice | |
|---|---|---|---|---|---|---|---|---|---|
| Lines heard | 3/3 | 3/3 | 3/3 | 3/3 | 3/3 | 3/3 | 1/3 | 3/3 | 3/3 |
| Voice | synth 0.16 | synth 0.151 | synth 0.107 | synth 0.128 | synth 0.087 | synth 0.072 | synth 0.076 | ours 0.958 | synth 0.233 |
| Motion | 3.61 | 3.01 | 1.77 | 3.82 | 1.54 | 1.72 | 0.73 | 1.70 | 1.65 |
| Scene changes | 0 | 0 | 0 | 3 | 0 | 0 | 0 | 2 | 0 |
| Loudness / peak | -20 / -4.2 | -15.3 / -2 | -21.7 / -7.3 | -11.6 / 0 | -20.1 / -6.4 | -19.2 / -4.3 | -14.6 / -4.4 | -13.9 / -0.2 | -17.5 / -2.4 |
| Length | 10.14s | 10.14s | 10.04s | 10.03s | 10.05s | 8.00s | 9.71s | 10.14s | 10.05s |
| Frame | 768×1344 | 768×1344 | 1072×1928 | 720×1280 | 720×1280 | 720×1280 | 576×1024 | 768×1344 | 720×1280 |
| Wait | 13s | 10:24 | 5:06 | 4:34 | 3:38 | 36s | 1:46 | 11:20 | 4:16 |
| Attempts | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| Cost | $0.80 | $0.76 | $1.96 | $1.00 | $1.80 | $3.20 | $0.45 | $0.76 | $1.80 |
The read
Same ranking on a different frame. H3 Max again: 13 seconds, every line, highest motion, identity held. Seedance 2.0 and H3 ref-to-video both kept the flyer hung from the lamp cord for the whole block, which is the beat; Kling and LTX were static around it. Wan cut three times inside a block that asked for none and mixed hot (−11.6 LUFS, 0.0 dBTP). Veo drew the bible's 'cloak' as large draped wings and delivered 8 seconds because that is its ceiling. Seedance's audio reference imitated the recording again (0.23), so that is a property of the model, not a one-off.
One person's judgment from the takes above, 2026-09-03. The table is the evidence; this is the opinion.
what whisper heard
- H3 Max: That's very kind, but then I'd be the wrong way up. Hmm. The jam is the problem. That is what I keep saying!
- H3 ref2v: that's very kind but then I'd be the wrong way up hmm the Jam is the problem that is what I keep saying
- Kling v3 Pro: That's very kind. But then I'd be the wrong way up. Mm-hmm. The jam is the problem. That is what I keep saying.
- Wan 2.7: That's very kind, but then I'd be the wrong way up The jam is the problem that is what I keep saying hmm
- Seedance 2.0: That's very kind, but then I'd be the wrong way up. The jam is the problem. That is what I keep saying.
- Veo 3.1: That's very kind. But then I'd be the wrong way up. Oh, the jam is the problem. That is what I keep saying.
- LTX-2: That's very kind, but then I'd be the wrong way up.
- H3 ref2v + voice: That's very kind, but then I'd be the wrong way up. The jam is the problem. That is what I keep saying.
- Seedance + voice: That's very kind. But then I'd be the wrong way up. Mmm, ba-ba. The jam is the problem. That is what I keep saying.