Offline capability manuals
Probe semantic controls under T2V, I2V, and R2V to estimate which requirements a frozen backend can satisfy.
NeurCo probes frozen video models to build model-specific capability manuals, then compiles user intent into execution strategies that keep reliable requirements and adapt the rest.
Instead of assuming a fixed instruction set, NeurCo measures what a generative renderer can reliably do, then compiles requests against that measured interface.
Probe semantic controls under T2V, I2V, and R2V to estimate which requirements a frozen backend can satisfy.
Represent intent as objects, constraints, events, and cameras, independent of any single video model.
Keep reliable requirements; rewrite, reroute, or invoke external operators when native execution is weak.
380 multi-requirement tasks with requirement-level scoring for compositional controllable generation.
NeurCo separates a shared semantic vocabulary from model-specific manuals, then lowers each request into backend-aware execution strategies.
Isolate controls at graded difficulties under applicable input modes.
Build a capability manual of reliable settings and preferred realizations.
Parse IR, preserve strong requirements, adapt weak ones with operators.
Run text, first-frame, reference, proxy, or shot-concatenation strategies.
Text only and Text+Operator cards compare against the original prompt, without an external prompt enhancer. External PE cards start from an externally enhanced prompt, then apply NeurCo on top. Hover a half to read that side's prompt (scroll if it is long). Wan reference-to-video runs are labeled R2V (VACE). Every clip stays at its original resolution; only cards on screen are decoded, so the full set can stay on the page without all of them playing at once.
Requirement satisfaction (Req.) and all-requirements success (All) on H3, Wan, and Seedance 2.5. NeurCo (T) is text-only compilation; NeurCo (T+O) adds external operators. VBench-2.0 testmini and Sci-VBench testmini are released as data, not shown as tables.
| Method | H3 | Wan | Seedance 2.5 | |||
|---|---|---|---|---|---|---|
| Req. | All | Req. | All | Req. | All | |
| Native inputs | ||||||
| Native | 66.5 | 28.2 | 46.8 | 8.4 | 68.5 | 26.9 |
| NeurCo (T) | 67.7 | 27.9 | 47.9 | 8.4 | 69.7 | 27.5 |
| NeurCo (T+O) | 69.3 | 25.3 | 52.3 | 11.8 | 68.2 | 25.8 |
| With external PE | ||||||
| External PE | 72.5 | 33.2 | 50.6 | 12.7 | 71.3 | 31.1 |
| External PE + NeurCo (O) | 73.0 | 33.7 | 52.5 | 14.5 | 72.1 | 32.2 |
| External PE + NeurCo (T) | 72.4 | 32.7 | 46.3 | 9.0 | 71.7 | 32.1 |
| External PE + NeurCo (T+O) | 72.0 | 34.5 | 50.0 | 12.3 | 71.8 | 31.5 |
On Seedance 2.5, text-only compilation is strongest on native inputs (68.5→69.7 Req.), while operator-augmented NeurCo helps more after external PE. Qualitative videos currently cover H3, Wan, and Seedance 2.5.
@article{neurco2026,
title = {Neural Rendering Compiler: Generative Video Models as Programmable Renderers},
author = {Author One and Author Two and Author Three and Author Four and Author Five},
journal = {arXiv preprint},
year = {2026}
}