{"schema_version":1,"research_id":"eoa_inverse_innovation_exp06_external_evaluation_20260803","source_assessment_id":"predictive_residual_processing__film_media_production:P4:v0","cell_id":"predictive_residual_processing__film_media_production","search_queries":["incremental rendering changed scene differential rendering VFX prior image research","VFX render farm bottleneck iteration rendering artist productivity official","USD Hydra change tracking dirty bits rendering official documentation","Foundry Katana live rendering ROI update official documentation","site:openusd.org Hydra dirty bits render index change tracking documentation","site:renderman.pixar.com incremental rendering checkpoint resume official","site:aws.amazon.com VFX rendering iteration render farm case study studio","VFX rendering compute costs artist iteration study render farm paper","\"prioritized re-rendering\" path traced scenes scene editing 2024","\"incremental updates\" path-traced scenes editing rerender image regions","selective re-rendering after scene changes indirect illumination paper","incremental rendering scene edits reuse previous frame path tracing","\"Residual path integrals for re-rendering\" SIGGRAPH project page","AWS Deadline Cloud pricing official render farm pricing","BLS software developers wage 2025 official occupational employment","MPA content security best practices VFX rendering official"],"sources":[{"source_id":"S1","title":"Scanline VFX managed 5M+ render hours through AWS Thinkbox Deadline for Aquaman","publisher":"Amazon Web Services","url":"https://aws.amazon.com/blogs/media/scanline-vfx-managed-5m-render-hours-through-aws-thinkbox-deadline-for-aquaman/","source_class":"COMMERCIAL_FIRST_PARTY","publication_date":"2019-01-02","accessed_at":"2026-08-03","claims_supported":["Professional VFX rendering can consume millions of render-hours and hundreds of thousands of tasks per day.","Scanline's head of engineering is an identifiable operational stakeholder responsible for render-farm capacity and workflow extensibility.","Large productions combine dedicated nodes, virtual nodes, and idle workstations, showing that render capacity is a material production resource."]},{"source_id":"S2","title":"Method Studios takes cloud rendering to the next level with AWS","publisher":"Amazon Web Services","url":"https://aws.amazon.com/blogs/media/method-studios-takes-cloud-rendering-to-the-next-level-with-aws/","source_class":"COMMERCIAL_FIRST_PARTY","publication_date":"2020-05-01","accessed_at":"2026-08-03","claims_supported":["Method Melbourne used 147,904 cloud-render tasks in four and a half weeks and scaled a 10,000-core internal farm fourfold.","Method's IT head expressly linked quicker render returns to iteration, creative process, and client artistic freedom.","Method identifies IT, systems, R&D, and VFX leadership as credible adopters or authorizers and reports completing a production-grade rendering proof of concept within a month.","The case study states that secure pipeline construction and pay-for-use rendering were material operational considerations."]},{"source_id":"S3","title":"Multiple Live Renders with Foresight+","publisher":"Foundry","url":"https://learn.foundry.com/katana/Content/ug/rendering_scene/render-foresight.html","source_class":"OFFICIAL_PRODUCT_DOCUMENTATION","publication_date":"n.d.","accessed_at":"2026-08-03","claims_supported":["Katana already dispatches multiple live renders locally or to a farm and propagates changes to individual render processes.","Only affected live renders need update, explicitly avoiding unnecessary render updates.","Katana provides synchronization and comparison controls for scene variables, live renders, and scene-graph geometry.","Affected-render updating is established commercial workflow prior art, although the documentation does not describe signed image residuals with independent full-render audits."]},{"source_id":"S4","title":"USD Terms and Concepts — Universal Scene Description 26.08 documentation","publisher":"Alliance for OpenUSD","url":"https://openusd.org/release/glossary.html?highlight=traverse","source_class":"STANDARD","publication_date":"2026-08","accessed_at":"2026-08-03","claims_supported":["Hydra is designed for large-scene change processing using scene delegates, a change-tracking render index, and render delegates.","Hydra is integrable with proprietary scene data and renderers and is used by Pixar for asset preview.","Versioned scene-change tracking needed by the proposal is technically available in an established production-oriented architecture."]},{"source_id":"S5","title":"Residual path integrals for re-rendering","publisher":"Bing Xu and research collaborators","url":"https://bingxu.tech/projects/2024_residual/page.html","source_class":"PRIMARY_RESEARCH","publication_date":"2024-07","accessed_at":"2026-08-03","claims_supported":["The research directly formulates the image difference between scene versions as a residual light-transport integral.","It targets incremental re-rendering when only a small portion of a scene changes and reports speedups over independent rendering.","The method covers moving objects and material editing, creating a substantial collision with the proposal's computational core.","The authors state that integration with camera-motion handling, path reuse, and residual-specific denoising is still needed for a shippable product."]},{"source_id":"S6","title":"Real-Time Editing of Path-Traced Scenes with Prioritized Re-Rendering","publisher":"SciTePress","url":"https://www.scitepress.org/PublishedPapers/2024/123246/","source_class":"PRIMARY_RESEARCH","publication_date":"2024","accessed_at":"2026-08-03","claims_supported":["The paper reports that minor scene edits commonly cause previous samples to be discarded and rendering to restart.","It implements adaptive incremental updates that identify, prioritize, and accumulate affected image regions.","Its user study found significant preference for incremental prioritized rendering over same-time full-screen re-rendering with denoising for small-object edits.","It supports feasibility and user value in interactive path tracing, but not production VFX fidelity, temporal effects, or net-cost claims."]},{"source_id":"S7","title":"Software Developers, Quality Assurance Analysts, and Testers","publisher":"U.S. Bureau of Labor Statistics","url":"https://www.bls.gov/ooh/Computer-and-Information-Technology/Software-developers.htm","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"2025","accessed_at":"2026-08-03","claims_supported":["The May 2024 median wage was $133,080 for software developers and $102,610 for software QA analysts and testers.","The published $63.20 combined hourly figure provides an official labor-cost anchor before benefits, overhead, specialist premiums, and 2026 escalation.","Pipeline development, testing, documentation, security requirements, and maintenance are material labor scopes."]},{"source_id":"S8","title":"AWS Deadline Cloud pricing","publisher":"Amazon Web Services","url":"https://aws.amazon.com/deadline-cloud/pricing/","source_class":"COMMERCIAL_FIRST_PARTY","publication_date":"2026","accessed_at":"2026-08-03","claims_supported":["Render cost depends on fleet type, instance size, duration, software licensing, storage, and transfer.","AWS provides a 400-frame example totaling $176.60 under stated Spot, Arnold, and storage assumptions and a 600-frame mixed-fleet example totaling $91.78 under different assumptions.","Prices are dynamic and regional, so proposal costs require measured workload assumptions rather than a universal per-frame rate.","A small retrospective test can purchase bounded compute without upfront commitment, while production costs may include licenses, storage, transfer, and connected-worker charges."]}],"problem_evidence":{"support":"MODERATE","rationale":"The problem is visible: S1 and S2 demonstrate that render capacity is operationally material at professional VFX scale, while S6 directly documents full restart after even minor path-traced scene edits. S2 also connects faster returns to creative iteration. However, no opened source measures what fraction of an offline VFX production's intermediate render load is unchanged, avoidable computation; the candidate's claimed prevalence and net burden therefore remain unquantified.","source_ids":["S1","S2","S6"]},"stakeholder_evidence":{"support":"STRONG","rationale":"Method's IT head expressly values quicker render returns for iteration and reports collaboration among IT, systems, R&D, and VFX leadership on a production-grade rendering proof of concept. Scanline's engineering head describes render management as integral and extensible per project. Foundry supplies a production product aimed at avoiding unnecessary render updates. These establish credible adopter and authorizer classes, although none commits to this particular audited-residual design.","source_ids":["S1","S2","S3"]},"prior_art":{"proximity":"SUBSTANTIAL_COLLISION","closest_analogues":[{"name":"Residual path integrals for re-rendering","similarity":"Directly represents the image difference between scene versions as a residual light-transport integral, focuses computation on nonzero residual contributions, and reports speedups versus independently rendering the new image.","remaining_difference":"The candidate adds production governance: checksum-bound scene and color provenance, random and risk-stratified complete-render audits, structured-error triggers, reviewer-decision testing, cumulative error budgets, and mandatory complete final renders. Those additions are proposed, not externally validated.","source_ids":["S5"]},{"name":"Prioritized incremental re-rendering","similarity":"Automatically identifies affected image regions, preserves prior accumulation elsewhere, prioritizes updates, and compares against conventional full-screen re-rendering in a user study.","remaining_difference":"The published work concerns interactive path-traced editing and eventual convergence; it does not establish an offline VFX version-reconstruction protocol, independent audit sampling, contractual final-render boundaries, or net production-resource savings.","source_ids":["S6"]},{"name":"Katana Foresight+ affected-render live updates","similarity":"Commercial production software routes scene changes only to affected live-render processes, supports farm dispatch, and synchronizes render-specific variables to avoid unnecessary updates and mismatches.","remaining_difference":"The documentation does not claim signed residual image reconstruction against a certified prior version, complete-frame shadow audits, consequence-weighted error budgets, or residual-driven fallback.","source_ids":["S3"]},{"name":"OpenUSD Hydra change processing","similarity":"Provides standardized large-scene change tracking and renderer integration that could identify dirty state and avoid unnecessary synchronization.","remaining_difference":"Hydra tracks scene changes; it does not by itself prove which pixels or light paths are reusable, generate image residuals, audit suppressed contributions, or certify review fidelity.","source_ids":["S4"]}],"distinctive_claim_remaining":"For intermediate VFX review, adding checksum-bound provenance, independent random and risk-stratified complete renders, structured spatial-temporal residual tests, human review, and automatic full-render fallback to existing incremental/residual rendering will reduce combined render, storage, transfer, and operator cost versus both complete rendering and ordinary affected-render/ROI practice without changing predefined reviewer decisions or concealing consequence-specific errors.","confidence":"HIGH"},"implementation_evidence":{"support":"MODERATE","rationale":"Scene-change tracking and affected-render dispatch are production-credible through Hydra and Katana, and two primary studies demonstrate incremental or residual path-traced rendering. The residual-path researchers nevertheless identify substantial work before a shippable product, and the sources do not validate deep compositing, simulations, stochastic seeds, denoising, motion blur, color-pipeline replay, multi-revision accumulation, or renderer-specific determinism. A read-only archived-shot test is legally and operationally bounded if authorized by the content owner; live approvals and final delivery remain excluded. Confidentiality, license, and contractual controls must be checked internally because no public source can establish rights for a particular shot.","source_ids":["S2","S3","S4","S5","S6"]},"scores":{"meaningful_impact":{"score":4,"rationale":"Render workloads can be extremely large and quicker returns have explicit creative value, but the avoidable share is unknown.","source_ids":["S1","S2","S6"]},"stakeholder_pull":{"score":4,"rationale":"Named studio engineering and IT stakeholders express strong need for scalable, faster, extensible rendering workflows; demand for this exact mechanism is unverified.","source_ids":["S1","S2","S3"]},"incremental_advantage":{"score":3,"rationale":"Audits, provenance, decision-level comparison, and fallback could add useful assurance beyond existing incremental rendering, but their net benefit has not been measured.","source_ids":["S3","S5","S6"]},"distinctiveness_plausibility":{"score":2,"rationale":"Residual image re-rendering and prioritized affected-region updating closely match the computational core; distinctiveness is mainly the integrated governance and audit layer.","source_ids":["S3","S5","S6"]},"technical_implementability":{"score":3,"rationale":"Key primitives exist, but the closest residual-rendering research is not yet a complete product and difficult temporal/global dependencies remain untested.","source_ids":["S3","S4","S5","S6"]},"adoption_authority_feasibility":{"score":4,"rationale":"VFX, pipeline, systems, R&D, and render-operations leadership are identifiable authorities, and an isolated proof of concept fits their demonstrated remit.","source_ids":["S1","S2"]},"evidence_readiness":{"score":2,"rationale":"Public evidence supports the problem class and mechanisms, but the decisive comparison requires proprietary archived shots, complete rerenders, and qualified reviewer judgments.","source_ids":["S2","S5","S6"]},"safety_net_benefit":{"score":4,"rationale":"Independent complete renders and mandatory fallback directly address plausible stale-image failures, though no source tests this safety architecture in VFX production.","source_ids":["S5","S6"]},"scalability":{"score":3,"rationale":"Render farms and change-processing architectures scale broadly, but audit cost, renderer integration, and fallback frequency could erase savings across complex shots.","source_ids":["S1","S3","S4","S8"]}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"10K_TO_50K","scope":"Preregister and run the proposed read-only retrospective shadow test on one completed shot, at most three archived transitions and 60 frames, including complete-render comparators, blinded review, instrumentation, and analysis.","confidence":"MODERATE","assumptions":["Approximately 120-400 combined hours of pipeline-TD, artist-review, render-operations, and analysis effort.","Existing renderer, farm, scene licenses, archived assets, and secure storage are available.","Compute remains bounded to the test; AWS examples show that raw render charges may be small relative to specialist labor, but scene complexity can vary greatly.","2026 resource-equivalent labor includes benefits, overhead, and specialist premiums above BLS cash wages."],"source_ids":["S7","S8"]},"initial_deployment_startup":{"band_2026_usd":"250K_TO_1M","scope":"Build a renderer-specific prototype with deterministic reconstruction, checksums, residual storage, audit selection, metrics, and fallback for one facility and one renderer.","confidence":"LOW","assumptions":["Roughly 1.5-4 specialist FTE-equivalents for 6-12 months.","Existing scene graph, render manager, storage, identity, and security systems can be integrated rather than replaced.","No final-delivery certification or multi-renderer support is included.","Specialist loaded labor substantially exceeds base wage because of benefits, overhead, and scarce VFX-pipeline expertise."],"source_ids":["S3","S4","S5","S7","S8"]},"operational_launch":{"band_2026_usd":"1M_TO_5M","scope":"Production-harden one facility across representative shot classes, including renderer and compositing integration, temporal/global-dependency tests, monitoring, security review, training, rollback drills, and parallel complete-render capacity.","confidence":"LOW","assumptions":["A cross-functional team of pipeline engineers, renderer specialists, QA, artists, operations, and security staff works for roughly 9-18 months.","Launch preserves existing complete-render infrastructure and initially duplicates some compute.","Licensing, storage, transfer, and render-farm usage vary by region, renderer, and fallback rate.","Multi-site, multi-renderer, or final-delivery use would exceed this scope."],"source_ids":["S2","S3","S5","S7","S8"]},"annual_recurring":{"band_2026_usd":"250K_TO_1M","scope":"Maintain one production deployment: 1-3 specialist FTE-equivalents, regression suites, audit renders, storage, monitoring, incident review, upgrades, and fallback capacity.","confidence":"LOW","assumptions":["A single facility and limited renderer set.","Audit and fallback rendering are a minority of total workload; otherwise recurring cost may exceed the band.","Existing farm management and security operations absorb shared overhead.","Compute, storage, data-transfer, and usage-based-license prices remain workload- and region-dependent."],"source_ids":["S7","S8"]}},"verified_pipeline_gates":{"externally_supported_problem":{"status":"YES","reason":"Large professional render workloads, explicit iteration value, and restart after small edits are externally documented, although offline redundant-work prevalence is not measured.","source_ids":["S1","S2","S6"]},"externally_credible_adopter_or_authorizer":{"status":"YES","reason":"Studio VFX, engineering, IT, systems, and R&D leadership are identifiable operational authorities with demonstrated rendering proof-of-concept and pipeline responsibilities.","source_ids":["S1","S2"]},"distinct_testable_incremental_claim":{"status":"YES","reason":"The candidate can be tested against full rendering and existing affected-render/ROI practice on resource use, consequence-specific error, fallback, and blinded reviewer-decision agreement.","source_ids":["S3","S5","S6"]},"bounded_next_evidence_step":{"status":"YES","reason":"A one-shot, three-transition, 60-frame retrospective test has bounded data, comparators, authority, stop rules, and measurable falsifiers.","source_ids":["S2","S5","S6","S8"]},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"For the bounded test only, content-owner authorization, read-only archived inputs, secure existing infrastructure, no live approval, no source mutation, preserved logs, and immediate complete-render fallback avoid an identified stop. Project-specific rights and security approval must still be obtained before execution.","source_ids":["S2"]},"credible_cost_scope_and_range":{"status":"YES","reason":"The ranges specify labor, integration, compute, storage, licensing, transfer, audit, and fallback assumptions and are anchored to official wage and render-pricing evidence; deployment estimates remain low-confidence because renderer and facility complexity are proprietary.","source_ids":["S7","S8"]}},"next_evidence_step":"With the archived-shot asset owner's written authorization, preregister and execute a read-only shadow test on one completed shot, no more than three version transitions and 60 frames. Compare (A) independent complete frame-range rendering, (B) the facility's ordinary cache/ROI/affected-render practice, and (C) checksum-bound residual reconstruction with random full-frame, boundary-tile, and risk-stratified temporal audits. Freeze renderer build, seeds, scene/assets, caches, color transform, resolution, review questions, spatial/color/temporal tolerances, sampling plan, fallback rules, and accounting rules before running. Qualified reviewers must answer identical predefined shot-note questions under blinded A/C presentation. Measure total compute, wall time, storage, transfer, setup and operator time, residual extent, structured audit error by dependency class, reviewer-decision disagreement, and fallback frequency. Falsify the incremental claim if C causes any concealed provenance mismatch or consequential miss, changes a predefined reviewer decision beyond the registered margin, fails deterministic replay, falls back above the registered ceiling, or does not reduce combined resource-equivalent cost by at least 20% versus both A and B with uncertainty intervals excluding zero. Stop immediately on source modification, rights/security breach, or failed checksum.","blocking_evidence":["No public evidence measures the proportion of real offline VFX revision compute that is predictable and safely reusable.","No opened source validates the integrated audit-and-fallback architecture on production-resolution VFX shots.","Net savings after prediction, residual rendering, complete-render audits, dual-path storage, operator time, and fallback are unknown.","Reviewer-decision equivalence and consequence-specific tolerances require qualified human evaluation on proprietary imagery.","Determinism and dependency completeness for simulations, deep compositing, global illumination, reflections, motion blur, denoising, random sampling, temporal operators, and color transforms are renderer- and shot-specific.","Project-specific content rights, confidentiality rules, software licenses, and authorization cannot be established by public web research.","World novelty, patentability, freedom to operate, market size, and realized impact remain unmeasured."],"research_disposition":"PARTNERED_RESEARCH_PROGRAM","world_novelty_boundary":"The search establishes substantial collisions in residual light-transport re-rendering, prioritized incremental image updates, affected-render live workflows, and scene-change tracking. It does not establish whether the exact combination of checksum-bound prior-image reconstruction, independent random and risk-stratified complete-render audits, structured spatial-temporal error triggers, blinded reviewer-decision comparison, and mandatory full-render fallback has appeared publicly. World novelty, patentability, freedom to operate, market size, and realized impact were not measured.","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_version":0,"controller_recommendation":{"action":"STOP_EMPIRICAL_RESEARCH_NEEDED","repairable":false,"material_progress_observed":true,"progress_targets":["Obtain written asset-owner, security, and pipeline authorization for a read-only archived-shot test.","Measure the full-render, ordinary cache/ROI, and audited-residual paths on the same preregistered transitions.","Demonstrate deterministic reconstruction with complete provenance and zero concealed version, cache, seed, or color-transform mismatches.","Show no consequential spatial, color, edge, reflection, simulation, denoising, or temporal miss in independent complete-render audits.","Show blinded reviewer-decision disagreement remains within the preregistered noninferiority margin.","Show at least 20% lower combined resource-equivalent cost than both comparators after labor, audit, storage, transfer, licensing, and fallback.","Estimate fallback frequency and identify shot/dependency classes where residual mode should be prohibited.","Differentiate the governance contribution from residual path integrals, prioritized re-rendering, Katana affected-render updates, and Hydra change processing."],"reason":"Bounded web research verifies the problem class, credible authorizers, implementable components, substantial prior-art collision, and a testable governance-level distinction. The remaining decision depends on proprietary archived scenes, complete rerenders, renderer determinism, internal rights, and qualified reviewer testing; it cannot be resolved by further bounded public-web research."},"proposal_index":4}