Sprint3D: Accelerating End-to-End 3D Asset Generation through Pipeline Optimization
Organizations: Alaya Lab · The University of Tokyo · University of California, Merced · National Yang Ming Chiao Tung University ·
Abstract
Image-to-3D generative models now produce asset-quality textured meshes, but at minutes per asset they remain slow to iterate with and costly to deploy. Prior work has mostly accelerated the sampling of the networks, and where it also optimizes the mesh processing, it trains its own generators. We instead start from a publicly released pipeline and ask how much of its end-to-end time can be saved. Profiling the public Pixal3D pipeline shows that its latency is dominated by what happens after the networks: the four rectified flows take 14% of the wall-clock time and mesh post-processing takes 85%, most of it in CPU-based UV unwrapping. We therefore first rebuild the mesh post-processing on the GPU and take it off the critical path, then cut the per-evaluation overhead of the flows and distill them from 76 to 8 network evaluations. The resulting pipeline, Sprint3D, reduces the median time per asset from 133 s to 4.05 s on one H200 GPU at the same export settings (33). At the same time, on 2,653 held-out Toys4K objects it improves every input-view normal metric of the Pixal3D protocol, with a slightly lower mean silhouette IoU, lower point-cloud alignment scores and less fine texture detail. The rebuilt post-processing alone also speeds up TRELLIS.2, which uses the same export code, by 6.7 without a detectable change in the evaluated metrics.
Figures & tables
| Optimization | Stage | Class | Effect (s) | Trade-off |
| (1) Mesh post-processing (Section 3.2 ) | ||||
| Tiled mesh path (16-tile atlas) | mesh post-proc. | 116.1 [ 141.3, 94.4] | 16-tile UV layout; superseded | |
| Mesh pre-processing on a worker thread | mesh texture | = | 0.20 [ 1.46, +1.08] | scheduling only |
| Fixed bake-kernel launch configuration | bake | = | 1.91 [ 3.19, 0.66] | identical in a unit test |
| Faster GPU simplification and charting | simplify, charts | 0.71 [ 2.10, +0.69] | non-deterministic, in spread | |
| GPU BVH build | BVH | 1.66 [ 2.79, 0.52] | ties change nearest face | |
| Critical path (s) | normal map | ||||||
| Configuration | Steps | E2E (s) | flows | post-processing | Speed-up | PSNR (full) | Mean ( ∘ ) |
| Base (Pixal3D) | 12/12/12/12 | 133.0 | 17.3 | 108.8 | – | reference | reference |
| Base (Pixal3D), re-run | 12/12/12/12 | 138.4 | 17.2 | 118.1 | – | [ , ] | [ , ] |
| + mesh path rebuilt | 12/12/12/12 | 19.0 | 17.2 | 1.32 | [5.85, 9.00] | [ , ] | [ , ] |
| + flow optimizations | 12/12/12/12 | 15.4 | 13.8 | 1.32 | [1.20, 1.27] | [ , ] | [ , ] |
| + distilled students ( Ours ) | 4/1/1/2 | 4.00 | 2.38 | 1.35 | [3.58, 4.31] | [ , ] | [ , ] |
| normal map | color | |||||
| Configuration | E2E (s) | Post-processing (s) | Speed-up | PSNR (full) | Mean ( ∘ ) | PSNR |
| TRELLIS.2 (official) | 112.7 | 91.4 | – | reference | reference | reference |
| + rebuilt mesh path, full-resolution remesh | 16.9 | 2.2 | [5.4, 8.6] | [ , ] | [ , ] | [ , ] |
| + adaptive half-resolution remesh | 16.5 | 2.1 | [5.4, 8.5] | [ , ] | [ , ] | [ , ] |
| Method | IoU (pt) | PSNR (overlap) | SSIM (overlap) | LPIPS (overlap) | Mean ( ∘ ) | Median ( ∘ ) | Mean B ( ∘ ) | 11.25 ∘ (pt) | 22.5 ∘ (pt) | 30 ∘ (pt) | E2E (s) |
| TRELLIS.2 | 79.55 | 17.15 | 0.577 | 0.033 | 25.06 | 18.97 | 27.56 | 48.27 | 64.47 | 70.80 | – |
| TripoSG | 59.87 | 14.22 | 0.504 | 0.051 | 38.87 | 34.31 | 40.88 | 22.74 | 35.89 | 42.98 | – |
| Hunyuan3D-2.1 | 59.82 | 14.31 | 0.505 | 0.049 | 39.33 | 35.24 | 41.80 | 23.20 | 36.10 | 43.17 | – |
| Base (Pixal3D) | 96.97 | 20.87 | 0.668 | 0.021 | 14.25 | 9.99 | 15.73 | 61.35 | 82.58 | 88.71 | 133.0 |
| Ours | 96.78 | 21.56 | 0.712 | 0.018 | 13.00 | 9.06 | 14.33 | 65.53 | 85.13 | 90.40 | 4.05 |
| Method | IoU (pt) | PSNR (full) | SSIM (full) | LPIPS (full) | Mean ( ∘ ) | Median ( ∘ ) | Mean B ( ∘ ) | 11.25 ∘ | 22.5 ∘ | 30 ∘ | CD ( ) | [email protected] | E2E (s) |
| Ours | 96.33 | 25.20 | 0.916 | 0.056 | 15.92 | 10.68 | 17.97 | 57.69 | 80.25 | 86.56 | 37.24 | 0.332 | 3.91 / 3.79 |
| Students replaced by alternatives | |||||||||||||
| Texture student (5k set) | 96.33 | 25.20 | 0.916 | 0.056 | 15.92 | 10.68 | 17.96 | 57.69 | 80.25 | 86.56 | 37.24 | 0.332 | 3.81 |
| 2-step shape students | 96.47 | 25.02 | 0.911 | 0.057 | 16.63 | 11.33 | 18.65 | 55.02 | 78.35 | 85.32 | 37.17 | 0.333 | 4.22 |
| Structure student (10k set) | 96.27 | 25.14 | 0.915 | 0.056 | 15.89 | 10.52 | 17.80 | 57.58 | 80.29 | 86.58 | 37.92 | 0.327 | 3.87 |
| Students replaced by their same-step teachers | |||||||||||||
| (a) Structure stage (400 views of 278 validation objects, true camera; support IoU at 32 3 ) | |||||||
| Method | Data | Steps | IoU | vs 12-step teacher [95% CI] | IoU vs 12-step teacher | GPU-h (+cache) | Updates |
| 12-step teacher | — | 12, guided | 0.597 | ref. | 1.000 | — | — |
| Consistency | 10k set | 4 | 0.596 | 0.002 [ 0.007, +0.003] | 0.797 | 27.8 | 12,000 |
| DMD2 | 10k set | 4 | 0.587 | 0.010 [ 0.015, 0.006] | 0.785 | 18.0 | 3,000 (15,000) |
| Trajectory (Ours) | 5k set | 4 | 0.598 | +0.001 [ 0.002, +0.004] | 0.894 | 0.25 (+3.5) | 500 |
| Stage (s, mean) | Base (Pixal3D) | Base (Pixal3D), re-run | + tiled mesh path | Ours |
| Input pre-processing | 0.12 | 0.23 | 0.23 | 0.22 |
| Structure flow | 1.12 | 1.12 | 1.13 | 0.35 |
| Coarse-shape flow | 1.59 | 1.64 | 1.65 | 0.19 |
| Fine-shape flow | 11.76 | 11.54 | 11.47 | 0.71 |
| Texture flow | 6.97 | 7.02 | 3.80 | 1.39 |
| Mesh extraction and remesh | 11.17 | 0.47 | 4.60 | 0.45 |
| E2E (s) | ||||||||||
| Configuration | PSNR (full) | SSIM (full) | Mean ( ∘ ) | CD ( ) | [email protected] | median | p90 | VRAM (GB) | Mean | PSNR |
| Base (Pixal3D), re-run | 24.27 | 0.9048 | 17.49 | 36.37 | 0.305 | 138.4 | 283.1 | 37.2 | [ , ] | [ , ] |
| + tiled mesh path (16-tile atlas) | 24.27 | 0.9048 | 17.47 | 36.40 | 0.304 | 55.3 | 82.0 | 29.1 | [ , ] | [ , ] |
| + mesh pre-processing on a worker thread | 24.28 | 0.9048 | 17.46 | 36.39 | 0.304 | 53.5 | 85.2 | 30.9 | [ , ] | [ , ] |
| + fixed bake-kernel launch configuration | 24.28 | 0.9048 | 17.46 | 36.40 | 0.304 | 52.8 | 81.8 | 31.3 | [ , ] | [ , ] |
| + faster GPU simplification and charting | 24.28 | 0.9048 | 17.46 | 36.42 | 0.304 | 52.4 | 81.4 | 31.1 | [ , ] | [ , ] |
| End-to-end time (s) | Median seconds (mean share) | |||||
| Method | Output | median | mean | p90 | sampling | post-processing |
| Base (Pixal3D) a | 1M faces, texture | 133.0 | 156.0 | 257.4 | 17.3 (16%) | 109.3 (83%) |
| TRELLIS.2 b | 1M faces, texture | 105.3 | 128.5 | 197.7 | 14.3 (16%) | 85.8 (84%) |
| Hunyuan3D-2.1 b,c | 40k faces, texture | 62.2 | 64.1 | 81.8 | 13.5 (22%) | 48.4 (78%) |
| TripoSG b | 1.2M faces, no texture | 7.9 | 7.8 | 10.6 | 2.7 (37%) | 4.9 (60%) |
| Method | IoU (pt) | PSNR (overlap) | SSIM (overlap) | LPIPS (overlap) | Mean ( ∘ ) | Median ( ∘ ) | Mean B ( ∘ ) | 11.25 ∘ (pt) | 22.5 ∘ (pt) | 30 ∘ (pt) | E2E (s) |
| (a) Validation set (400 objects, one view each) | |||||||||||
| TRELLIS.2 | 71.88 | 15.80 | 0.521 | 0.084 | 28.92 | 21.07 | 34.00 | 42.13 | 56.97 | 64.18 | – |
| TripoSG | 60.35 | 13.39 | 0.477 | 0.105 | 40.45 | 34.57 | 45.59 | 17.18 | 34.95 | 45.35 | – |
| Hunyuan3D-2.1 | 64.90 | 14.47 | 0.499 | 0.093 | 34.56 | 27.87 | 40.75 | 28.47 | 45.11 | 54.56 | – |
| Base (Pixal3D) | 96.60 | 19.20 | 0.613 | 0.053 | 17.22 | 11.68 | 19.42 | 53.36 | 77.32 | 84.63 | 133.0 |
| Ours | 96.33 | 19.82 | 0.655 | 0.048 | 15.92 | 10.68 | 17.96 | 57.69 | 80.25 | 86.55 | 4.05 |
| Normal map | Single-view 3D | ||||||
| Method | PSNR (full) | SSIM (full) | LPIPS (full) | CD ( ) | [email protected] | [email protected] | [email protected] |
| TRELLIS.2 | 23.15 | 0.929 | 0.076 | 28.01 | 0.302 | 0.523 | 0.712 |
| TripoSG | 19.52 | 0.908 | 0.159 | 58.62 | 0.124 | 0.239 | 0.381 |
| Hunyuan3D-2.1 | 19.76 | 0.910 | 0.158 | 59.94 | 0.143 | 0.264 | 0.400 |
| Base (Pixal3D) | 28.86 | 0.956 | 0.024 | 26.61 | 0.207 | 0.392 | 0.598 |
| Ours | 29.27 ∗ | 0.961 ∗ | 0.022 ∗ | 25.55 ∗ | 0.220 ∗ | 0.408 ∗ | 0.614 ∗ |
| Metric | (Ours Base (Pixal3D)) | 95% CI | better / worse / tie | |
| Input-view normal maps (Holm over 13) | ||||
| IoU (pt) ∗ | [ , ] | 1555 / 1095 / 1 | ||
| PSNR (overlap) ∗ | [ , ] | 2212 / 439 / 0 | ||
| SSIM (overlap) ∗ | [ , ] | 2268 / 383 / 0 | ||
| LPIPS (overlap) ∗ | [ , ] | 1975 / 676 / 0 | ||
| PSNR (full) ∗ | [ , ] | 1956 / 695 / 0 | ||
| Geometry only | Colored | |||
| Method | Uni3D | ULIP-2 | Uni3D | ULIP-2 |
| TRELLIS.2 | 30.70 | 31.38 | 33.34 | 36.67 |
| TripoSG | 30.42 | 31.43 | 30.42 | 31.43 |
| Hunyuan3D-2.1 | 26.88 | 26.88 | 29.88 | 32.46 |
| Base (Pixal3D) | 29.88 | 30.23 | 32.88 | 35.64 |
| Ours | 29.14 | 29.17 | 32.58 | 34.94 |
| Method | CD- ( ) | CD- ( ) | EMD ( ) | [email protected] | [email protected] |
| (a) Validation set (400 objects, one view each) | |||||
| TRELLIS.2 | 1.87 | 2.85 | 9.92 | 0.521 | 0.878 |
| TripoSG | 3.89 | 4.10 | 12.53 | 0.383 | 0.761 |
| Hunyuan3D-2.1 | 2.11 | 3.03 | 10.59 | 0.500 | 0.858 |
| Base (Pixal3D) | 3.65 | 3.77 | 11.59 | 0.447 | 0.796 |
| Ours | 3.42 | 3.67 | 11.34 | 0.455 | 0.803 |
| Input view | All views | Distance to GT renders | |||||||
| Method | PSNR | SSIM | LPIPS | PSNR | SSIM | LPIPS | FD (Inception) | FD (DINOv2) | CMMD |
| Base (Pixal3D) | 24.34 | 0.915 | 0.066 | 16.50 | 0.829 | 0.233 | 32.37 | 292.8 | 0.0557 |
| Ours | 24.30 | 0.920 ∗ | 0.061 ∗ | 16.58 ∗ | 0.835 ∗ | 0.225 ∗ | 32.68 | 313.0 † | 0.0624 |
| 12-step teacher | 24.21 | 0.920 | 0.061 | 16.59 | 0.834 | 0.226 | 31.75 | 293.5 | 0.0421 |
| Difference | 0.000 | 0.000 | 0.000 | 0.000 | |||||
| 95% CI | [ , ] | [ , ] | [ , ] | [ , ] | [ , ] | [ , 0.000] | [ , ] | [ , ] | [ , ] |
| Structure stage (400 views of 278 validation objects, true camera): support | |||||||||
| Method | Data | Updates | Sampler | IoU | P | R | Thin R | IoU vs T | GPU-h |
| 12-step teacher | — | — | Euler | 0.597 | 0.725 | 0.727 | 0.476 | 1.000 | — |
| 25-step teacher | — | — | Euler | 0.597 | 0.724 | 0.727 | 0.472 | 0.973 | — |
| 12-step teacher, shift 5.0 | — | — | Euler | 0.598 | 0.725 | 0.727 | 0.480 | 0.978 | — |
| Consistency | 5k (4,993) | 15,000 | re-noise | 0.592 | 0.724 | 0.718 | 0.460 | 0.796 | n/r (16) |
| Consistency | 10k (19,998) | 12,000 | re-noise | 0.596 | 0.728 | 0.722 | 0.464 | 0.797 | 27.8 (4) |
3D Asset Generation
3D asset generation creates and edits objects, characters, materials, and scenes from text, images, video, or geometric constraints.
See all 176 papers