Two people, two lines
Two speakers in one frame: the right mouth has to move on the right line, and a pizza box has to survive being held.
2 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, H3 ref2v, Wan 2.7, LTX-2
The woman on the wall looks up at the man and says: "You're late." The man lifts the pizza box a little and says: "I brought pizza." She uncrosses her arms and laughs. The camera pushes in slowly. Ambient city sound, no music.
Kling v3 Pro
[{"prompt":"The woman on the wall looks up at the man and says: \"You're late.\" The man lifts the pizza box a little and says: \"I brought pizza.\" She uncrosses her arms and laughs. The camera pushes in slowly. Ambient city sound, no music.","duration":"5"}]- “You're late.”
- “I brought pizza.”
The images or videos provided may contain likenesses of real people or other private information that cannot be processed."
| H3 Max | H3 ref2v | Kling v3 Pro | Wan 2.7 | Seedance 2.0 | LTX-2 | |
|---|---|---|---|---|---|---|
| Lines heard | 2/2 | 2/2 | 2/2 | 2/2 | — | 2/2 |
| Voice | model's own | model's own | model's own | model's own | — | model's own |
| Motion | 2.36 | 3.24 | 1.28 | 1.02 | — | 1.12 |
| Scene changes | 0 | 0 | 0 | 1 | — | 0 |
| Loudness / peak | -16.3 / -2.7 | -12.3 / -0.1 | -19 / -5.4 | -10.3 / 0 | — | -19.9 / -5.2 |
| Length | 5.18s | 5.18s | 5.04s | 5.04s | — | 4.92s |
| Frame | 768×1344 | 768×1344 | 1072×1928 | 720×1280 | — | 576×1024 |
| Wait | 6s | 2:40 | 3:36 | 2:32 | — | 39s |
| Attempts | 1 | 1 | 1 | 1 | — | 1 |
| Cost | $0.40 | $0.30 | $0.98 | $0.50 | — | $0.22 |
The read
Two lines, two speakers, and a pizza box. H3 Max continued the frame with a slow push-in, kept both faces, and the box stayed a box; Kling held the composition and both lines but barely moved anyone. H3 reference-to-video re-staged the rooftop — the man standing further back, the box lifted high — with better line delivery and worse continuity with the frame, which is the same trade it made on Night Kitchen. Wan opened on the man's back, then cut to a new angle halfway through a five-second clip that asked for none; the scene-change count says so. LTX re-framed to the woman alone, lost the box, and heard both lines anyway, which is why the transcript check is not a quality score. Seedance refused the frame outright as a possible likeness of real people — the one result here that is a fact about the model rather than the take.
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: You're late. I brought pizza.
- H3 ref2v: You're late. I brought pizza.
- Kling v3 Pro: You're late. I brought pizza. [ Laughs ]
- Wan 2.7: You're late. I brought pizza.
- LTX-2: You're late I Brought pizza