A dance can cross styles. Details don’t always follow.
Two useful motion-transfer examples. Neither is an exact identity or whole-scene conversion.
Real performance → drawn character
Compare Real performance → drawn character
Inspect movement closely
Both videos stay on the same timeline at every speed; this does not retime the character’s moves. Frame buttons seek by 41.7 ms on the output grid; browser decoding and a different source frame rate can affect the displayed frame.
Loading the comparison videos… Open the result file
Silent publication previews. This pair compares the shared interval; full-length files are below. Audio findings refer to preserved original test files.
What to look for The dance is recognizable, but the stocky ink character becomes slimmer and more 3D-like. Starting and ending poses pull toward the appearance reference.
Original video, full results & downloads
Drawn action → photographic character
Compare Drawn action → photographic character
Inspect movement closely
Both videos stay on the same timeline at every speed; this does not retime the character’s moves. Frame buttons seek by 41.7 ms on the output grid; browser decoding and a different source frame rate can affect the displayed frame.
Loading the comparison videos… Open the result file
Silent publication previews. This pair compares the shared interval; full-length files are below. Audio findings refer to preserved original test files.
What to look for Guard, crouch, palm strike and return carry through. Face identity, muscle proportions and the cartoon energy effect do not.
Original video, full results & downloads
AI-generated outputs are identified in each comparison. Public previews omit source audio where present; visual streams are unchanged. Original test files and audio measurements remain preserved. No performer or photographer endorses ModelPilot.
The finding
Can a real performance drive a drawn character, and vice versa?
Both directions worked well enough to demonstrate the action. The drawn dancer became slimmer and more smoothly shaded; the realistic fighter gained muscle and lost the reference face and energy effect.
One completed example in each direction, within a six-render campaign. Sampled visual review; no success-rate estimate.
What we would choose
Offer this as motion-guided character reinterpretation. It is useful for exploration and animation ideas, with review required before using the result as a final character asset.
Where it still falls short
- The background does not become a matching cartoon scene just because the character is drawn.
- A single example in each direction cannot establish reliability across animation styles.
- Faces, body proportions, hands, endpoint poses and effects can diverge from the reference.
What changed
| Choice | Baseline / first option | Candidate / second option |
|---|---|---|
| Real → drawn | Live-action studio dance | Iron Vagrant appearance reference |
| Drawn → real | Cartoon palm-strike action | Photographic athletic reference |
| Control type | Source motion vs result | Illustrative examples, not paired improvements |
Time & cost, without the guesswork.
| Saved run | Recorded processing | Recorded amount | Scope |
|---|---|---|---|
| Real → drawn | 16m 05s | $0.7396 | Allocated worker share; queue 12m 25s |
| Drawn → real | 6m 01s | $0.2760 | Allocated worker share; queue 30m 23s |
Per-video amounts are time-weighted shares of the campaign’s actual $3.237 worker charge, not separate invoiced per-video prices. Total campaign included six different tests. Historical observations only.
The saved smaller-reference option targets RTX PRO 6000 (96GB). These are not controlled hardware benchmarks.
Inspect the setup
- Recipe family
- H3 pose-guided motion transfer, smaller motion-reference candidate
- Roles
- The video supplies action; the still supplies appearance. Neither should be read as an exact identity guarantee.
- Seed coverage
- One sampled render per direction
- What stayed unresolved
- Broad style reliability, effect transfer and production-ready character consistency
- Hardware and timing limits
- The saved smaller-reference option targets RTX PRO 6000 (96GB). These are not controlled hardware benchmarks.
Study settings summary; not an importable ComfyUI workflow. Requires the private ModelPilot runtime/nodes and separately licensed model weights.
Download study filesIncludes recorded settings, media hashes, credits and license notices. A complete executable graph is not included for this study.
Sources, attribution & publication status
Source credits, original licenses and adaptation notices are recorded below. Results are AI-generated research demonstrations, not endorsements. Model use and output redistribution remain subject to the MiniMax H3 license and any required separate authorization.
All videos on this page are saved outputs, not new generations. Settings and conclusions were transcribed from the original experiment receipts. No customer accounts, private batch identifiers or credentials are included in the downloadable summary.
- Studio dance source — Andres Daza. Prepared and excerpted for the saved test; no performer endorsement. License recorded at acquisition.
- Iron Vagrant and photographic character references — Rising Heaven / ModelPilot generated test assets. Existing saved references and generated palm-strike source; no new generation.
Source credits and changes accompany each download. Read the media notices. Public previews omit source audio; the original test files remain preserved. ModelPilot has received authorization for its H3 use and works. This does not grant rights to third-party source material.