What do studio and vendor first-person accounts say about the real speed and accuracy gains of ML deformers in production rigs, and where do they fall short?

DreamWorks and Berkeley (2018) report 5x-10x faster deformation evaluation on production rigs. Golaem (DigiPro 2024) tried to reproduce this in a commercial crowd product and reports only x1.2-x2 on production assets because the paper's rigs were optimised for the model and the rest of rig evaluation dominates; its main disappointment was no significant speed gain. It reports Euler-angle and facial limits of the first method, ships the ML deformer as a crowd-package feature for close-up facial deformation, and notes crowd artists prefer overfitting because crowd sims blend clips rather than generate unseen animation. The paper's own Table 1 fps values imply 1.8x-2.5x, which does not match its text range. Curator qualifications: Bailey: 'reduces the computational time spent on evaluating deformations by a factor of 5×-10×'; notice '... posted here for your personal use. Not for redistribution. The definitive Version of ...'. Golaem: 'we could only reach a x1.2-x2 speed up with production assets'; 'a potential gain of maximum factor 2 in rig evaluation speed is not sufficient to justify a new product'; Table 1 CharA 8.7 to 22.1 fps (2.54x), CharB 10.1 to 18.6 fps (1.84x), Maya 2022 parallel evaluation, 5000 frames, RTX 4060, 16 CPUs; Table 2 per-frame deformer times about 12-19 ms across backends ('none of the implemented optimization presents a significant improvement'). CharA's Table 1 ratio exceeds the text's 'maximum factor 2'. ADDENDUM-001 (A2) label fix (S15 -> S18 in LEAD-QA) applies; candidate itself already cites S18. Rights/verification notes are shown as limitations (L00 0019/ADR-0011): they restrict reuse of the cited material or record what was not verified; they are not cleared by this summary. Limitations: [rights_conflict_record] Bailey 2018 author version is 'Not for redistribution': cite/link the version of record; no excerpt beyond short quotation (rights_conflicts_for_curation preserved). Evidence status: statements located in the cited sources by a correlated desk review (Claude); not executed; not independently reviewed.

{
  "citations": [
    {
      "url": "https://research.dreamworks.com/wp-content/uploads/2018/08/Rig_Approximation-Edited.pdf",
      "note": "Bailey et al. 2018 author-version PDF (equals producer pin): abstract line 24; notice line 32; performance section; arXiv 2406.09783v1 (DigiPro 2024) pdftotext -layout: Sec. 6.4 lines 279-302, Table 1 lines 253-256, Table 2 lines 260-267"
    },
    {
      "url": "https://arxiv.org/abs/2406.09783v1",
      "note": "Bailey et al. 2018 author-version PDF (equals producer pin): abstract line 24; notice line 32; performance section; arXiv 2406.09783v1 (DigiPro 2024) pdftotext -layout: Sec. 6.4 lines 279-302, Table 1 lines 253-256, Table 2 lines 260-267"
    },
    {
      "url": "http://web.archive.org/web/20230126203603id_/https://www.unrealengine.com/en-US/blog/battle-testing-unreal-engine-5-1-s-new-features-on-fortnite-battle-royale-chapter-4",
      "note": "Bailey et al. 2018 author-version PDF (equals producer pin): abstract line 24; notice line 32; performance section; arXiv 2406.09783v1 (DigiPro 2024) pdftotext -layout: Sec. 6.4 lines 279-302, Table 1 lines 253-256, Table 2 lines 260-267"
    }
  ]
}