Curvelle Kling tests

Test board for the Curvelle cartoon template (Amua-style lip-synced song cartoons). Each round adds cards here. Template and evidence: knowledge/brands/curvelle/kling-cartoon-template-2026-10-17.md, output/cartoon-style-gap-2026-10-17/FINDINGS.md.

175 crspent on tests so far
720p · 9:16all tests

Round 1 · Calibration (2026-10-17)

Price each Kling route per second and test lip-sync to a supplied song slice. 4 × 5 s jobs, 720p, run one at a time with the credit balance read before and after each. Result: keyframe route wins for story shots (9 cr/s); Kling does not lip-sync a supplied song, so singing shots move to sync-3.

T1-sing-omni-audio

omni_ref_video · kling-video-v4_0 · 5 s · 720p

Cost
45.0 cr
Rate
9 cr/s
Gen time
70.2 s
Status
COMPLETED

Two-character still + English prompt. Kling kept our song slice on the track, but the mouth only moves loosely (both coders: loose/none). Not usable for lip-sync.

pro 4/10 · lip-sync loose · Mouth movements are generic and do not match the rap audio.

flash 7/10 · lip-sync loose · Solid Pixar-like styling and legible branding with loose rap lip-syncing.

frame_anim_4.png
s1_slice.mp3
Prompt

[STYLE] The young woman in 图片1 sings 音频1 directly to camera with clear, accurate lip-sync, every word on her lips, confident sassy expression, small shoulder sway to the beat, slow push-in. Her mouth matches 音频1 throughout.

T1b-sing-omni-audio-cn

omni_ref_video · kling-video-v4_0 · 5 s · 720p

Cost
45.0 cr
Rate
9 cr/s
Gen time
93.5 s
Status
COMPLETED

Retry: single-character sheet + Kling's documented phrasing (the character in image 1 speaks following audio 1). Kling IGNORED our song and invented new rap lyrics ('watch me now, I'm taking it higher'). audio_1 is a voice/style reference, not a lip-sync driver → SING shots move to sync-3 lip-sync.

pro 3/10 · lip-sync none · Audio differs from prompt, so lip-sync does not match the requested lyrics.

flash 3/10 · lip-sync none · Audio spoken does not match the prompt's target lyrics, resulting in failed lip-sync.

anim_nadia_after.png
s1_slice.mp3
Prompt

图片1的角色按音频1说话,嘴型与音频1逐字同步。[STYLE] Only one character on screen: the young woman from 图片1, medium close-up, facing camera, rapping 音频1 with clear, accurate lip-sync, confident expression, slow push-in.

T2-act-v4-keyframes

image_to_video · kling-video-v4_0 · 5 s · 720p

Cost
45.0 cr
Rate
9 cr/s
Gen time
199.1 s
Status
COMPLETED

Winner for ACT shots. One job = 3 story shots with clean hard cuts at the keyframes, label reads CURVELLE (both coders 7/10). 9 cr/s, slowest (199 s).

pro 7/10 · lip-sync n/a · All shots appear with clean cuts and match the requested style.

flash 7/10 · lip-sync n/a · All three shots cut cleanly with consistent Pixar styling, accurately matching the intended sequence and bottle label.

first_image: frame_anim_1.png
keyframe_1: frame_anim_2.png
tail_image: frame_anim_3.png
Prompt

[STYLE] Three shots with hard cuts at each keyframe. 0-1.7 s: the seamstress tugs the loose fabric at the young woman's hip. 1.7-3.4 s: in a kitchen at night she pokes her small belly and slumps. 3.4-5 s: the grandmother slides the CURVELLE dropper bottle across the table and winks. No talking, mouths closed.

T3-act-v3turbo

image_to_video · kling-video-v3_0_turbo · 5 s · 720p

Cost
40.0 cr
Rate
8 cr/s
Gen time
109.2 s
Status
COMPLETED

Single-image twirl. 8 cr/s, only ~11% cheaper than v4, and the seamstress reaction is weaker (coders 5 vs 7). Not worth routing to.

pro 5/10 · lip-sync none · Woman turns, but seamstress glasses remain fixed.

flash 7/10 · lip-sync n/a · The twirl and seamstress reaction execute cleanly with strong Pixar styling and readable product labeling.

frame_anim_4.png
Prompt

[STYLE] The young woman does a confident slow twirl in the fitted pink gown while the seamstress steps back amazed, glasses sliding down her nose. No talking, mouths closed.

[STYLE] block used in every prompt

Vertical 9:16 3D animated film in the Pixar/Disney style: big expressive eyes, smooth stylized skin, warm saturated color, cinematic contrast lighting, rich detailed sets, slow push-in camera. Women have an exaggerated hourglass figure: small waist, flat stomach, wide hips, full thighs, in fitted clothes that show the shape. No subtitles, no captions, no on-screen text except the word CURVELLE on the bottle label.