AIFD / PUBLIC STANDARD AIFD-STD-DEF-001 AI Film Definitions & Classification DÉFINITIONS ET CLASSIFICATION DES FILMS IA STANDARD EDITION 5.1 FIRST PUBLISHED 1 August 2024 EDITION ADOPTED 30 June 2026 English text authoritative. French titles support international navigation. © 2024-2026 International Federation of AI Film Producers’ Associations. Reproduction for conformity, review and public accountability requires attribution and preservation of the requirements. “Shall” expresses a requirement; “should” expresses a recommendation. ifaifpa.org / standards Verify records: ifaifpa.org/registry AIFD-STD-DEF-001 / 5.1 AIFD-STD-DEF-001 2 Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 — Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 0 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2 Normative references . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 3 Terms and definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 4 The four axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 5 Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 6 Declaration and labelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 7 Misclassification and review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Annex A (normative) — Axis determination matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Annex B (informative) — Worked examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Annex R Revision history . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Bibliography (informative) AIFD-STD-DEF-001 / 5.1 AIFD-STD-DEF-001 3 — Foreword The International Federation of AI Film Producers’ Associations (IFAIFPA) publishes the AIFD catalogue of standards for the disclosure, identity and accountability of AI film production, and operates the accreditation schemes that assess conformity with them. Standards are drafted by the federation's standards function, exposed to public consultation, and adopted by resolution of the governing body. Requirements use the verb shall; recommendations use should; permissions use may. Notes and informative annexes do not contain requirements. This document, AIFD-STD-DEF-001 AI Film Definitions & Classification, was first published on 1 August 2024. The present edition was adopted on 30 June 2026. The revision history in the final annex records every edition and its principal changes. The English text is authoritative; le texte anglais fait foi. Nothing in this document grants any right to use a federation mark, and no conformity claim may be made except through the assessment and decision process defined in AIFD-STD-REV-001 and recorded in the federation's public register. 0 Introduction Film works now reach audiences through production pipelines in which artificial-intelligence systems may contribute anything from noise reduction to entire performances. Festivals, distributors, insurers, journalists and audiences all ask the same first question — what, in this work, was made by an AI system? — and receive answers in vocabularies that do not match. This standard exists so that one vocabulary, with defined terms and reproducible determination rules, underlies every declaration, entry condition, register record and certificate in the AIFD system. It deliberately classifies means of production, never quality: an AI-Native work is not better or worse than an AI-Assisted one; it is differently made, and the difference is a fact the public is entitled to know. 1 Scope This standard defines the terms used across the AIFD catalogue and establishes the classification of film works by their use of AI systems. It applies to: (a) any work declared, entered, screened, registered or assessed under an IFAIFPA scheme; (b) any festival, organisation or person making a classification claim that references AIFD; and (c) the machine-readable representation of such claims. This standard does not apply to the assessment of artistic quality, does not create rights in any work, and does not by itself authorise any conformity claim. Classification under this standard is a statement of fact about production means. AIFD-STD-DEF-001 / 5.1 AIFD-STD-DEF-001 4 2 Normative references The following documents are indispensable for the application of this standard. For dated references, only the edition cited applies; for undated references, the latest edition applies. AIFD-STD-PRO-001, Production Disclosure & Accountability AIFD-STD-RGT-001, Rights, Provenance & Identity AIFD-STD-REV-001, Independent Review, Complaints & Appeals C2PA Specification (as referenced through AIFD-STD-RGT-001) 3 Terms and definitions 3.1 AI system Machine-based system that, for explicit or implicit objectives, generates outputs such as images, video, audio, text or decisions that influence the content of a film work. 3.2 generative model AI system component trained on data and capable of producing novel perceptible content in response to prompts, conditioning signals or other inputs. 3.3 generative contribution Element of the finished work whose perceptible content was produced by an AI system, whether or not subsequently edited by a person. 3.4 assistive use Use of an AI system for support tasks that do not by themselves place perceptible content in the finished work. EXAMPLES Rotoscoping assistance; automated dialogue transcription; colour-match suggestions accepted or rejected shot-by-shot by an operator; translation drafts revised by a person. 3.5 material use Use of an AI system such that generative contribution is perceptibly present in the finished work on the axis concerned, interleaved with recorded or authored material. 3.6 constitutive use Use of an AI system such that the axis concerned substantially consists of generative contribution. AIFD-STD-DEF-001 / 5.1 AIFD-STD-DEF-001 5 3.7 synthetic performer Depicted performer whose likeness, movement or voice does not correspond, in whole or in part, to a recorded performance by the natural person depicted, including wholly invented persons. 3.8 digital replica Synthetic performer intended to be recognised as a specific identifiable natural person, living or deceased. 3.9 recorded performance Performance captured from a natural person by camera, microphone, motion capture or equivalent means, including volumetric capture. 3.10 axis One of the four assessed dimensions of a work: image, script, performance, sound; each includes the post-production operations attributable to it. 3.11 finished work The version of the work as exhibited or distributed, fixed by content hash under AIFD-STD-RGT-001. 3.12 version Any cut, edit, localisation or remaster of a work distributed under the same or a related title. 3.13 declarant Responsible producer (AIFD-STD-PRO-001) or, for festival-side claims, the festival operator making a classification statement. 3.14 classification Assignment of a work to one of the classes in 5.3 on the basis of axis determinations under clause 5. 4 The four axes 4.1 Image The image axis covers all visual content of the finished work: photography, animation, compositing, set extension, crowd replication, de-aging, style transfer, upscaling and every other operation whose result is visible. Restorative operations that recover a signal already present (denoising, stabilisation, scratch removal) are assistive unless they synthesise content absent from the source. AIFD-STD-DEF-001 / 5.1 AIFD-STD-DEF-001 6 4.2 Script The script axis covers story, structure, dialogue, narration and on-screen text. Machine translation of a completed human script for subtitling is assistive; machine-generated dialogue or story structure retained in the finished work is material or constitutive according to its share. 4.3 Performance The performance axis covers the depiction of performers: face, body, movement and the acted voice. Motion retargeting of a recorded performance onto a character is assistive where the performance remains that of the recorded person; synthesis of face, body or voice beyond a recorded performance is material at minimum; a synthetic performer central to the work renders the axis constitutive (see 5.2). 4.4 Sound The sound axis covers music, effects, ambience, mix and the non-acted voice (narration is assessed under both script, for its text, and sound, for its voice). AI-composed score retained in the finished work is material or constitutive according to its share; assistive examples include stem separation, loudness conformance and noise reduction. 5 Classification 5.1 Determination per axis The declarant shall determine, for each axis, the highest level of AI use present in the finished work: none, assistive (3.4), material (3.5) or constitutive (3.6). The determination follows the finished work as exhibited, not the production process in the abstract: generative material created but not retained does not count; retained material counts however small the tool's role appeared in the pipeline. 5.2 Fixed minima The following minima apply regardless of the declarant's assessment: (a) a synthetic performer on screen renders the performance axis at minimum material, and constitutive where the synthetic performer is a principal character; (b) a digital replica additionally engages AIFD-STD-RGT-001 clause 5; (c) machine-generated dialogue exceeding incidental lines renders the script axis at minimum material. 5.3 Classes The work's class follows from its axis determinations: CLASS CONDITION AI-Native At least two axes constitutive, or the image axis alone constitutive. AI-Hybrid At least one axis material or constitutive, and the work is not AI-Native. AIFD-STD-DEF-001 / 5.1 AIFD-STD-DEF-001 7 CLASS CONDITION AI-Assisted No axis exceeds assistive. (outside classes) No AI contribution on any axis. The work shall not carry an AIFD class label. 5.4 Undetermined axes Where a declarant cannot determine an axis in good faith after reasonable inquiry — typically for licensed archive or third-party material — the axis shall be declared undetermined, and the work takes the class that would result from the least favourable plausible determination of that axis. An undetermined axis shall be re-determined when the missing information becomes available. 5.5 Special cases Anthology and episodic works are classified per constituent part, and as a whole by the highest class present among parts. Restorations and re-releases are classified by the operations applied in the new version, not the original production. Documentary works classify their non-archival content normally; archival material is assessed for synthetic alteration only. Live and interactive works are classified by the rules of the exhibited session; where content varies per session, the declaration states the generating configuration. Trailers, key art and promotional material carry their own classification when they materially differ from the work. 5.6 Re-determination A new version (3.12) is a new determination. Classification travels with the version's content hash; exhibiting a differently-classified version under the label of another version is misclassification under clause 7. 6 Declaration and labelling 6.1 Content of a class claim Every classification claim referencing AIFD shall state: the class; the edition of this standard applied; the identity of the declarant; and the content hash of the version concerned where a hash exists under AIFD-STD-RGT-001. Axis-level determinations shall be available to any scheme or accredited festival receiving the claim, and should be published with the work's disclosure summary. AIFD-STD-DEF-001 / 5.1 AIFD-STD-DEF-001 8 6.2 Machine-readable form Claims shall be reproducible on the federation's published disclosure template, carrying at minimum the fields of Annex A of AIFD-STD-PRO-001 relevant to classification. A human-readable label shall quote the class name exactly as defined in 5.3 and shall not add qualifiers that could be read as federation endorsement. 6.3 Use by festivals Festival entry conditions that reference classes shall quote them exactly and apply them as published. A festival shall not redefine a class term while citing this standard. 7 Misclassification and review Classification disputes concerning works entered under an accredited scheme are handled under AIFD-STD-REV-001. A revised classification replaces the previous one prospectively; the register of decisions preserves history. Deliberate or reckless misclassification by a declarant is a critical non-conformity under AIFD-STD-PRO-001 clause 10; festival tolerance of known misclassification engages gate G5 of AIFD-STD-FES-001. Annex A (normative) — Axis determination matrix Reviewers and declarants apply the matrix below when a level is contested. The level of an axis is the highest row matched by any retained element. LEVEL IMAGE SCRIPT PERFORMANCE SOUND Assistive Restoration, stabilisation, technical upscaling of existing signal Translation drafts, transcription, continuity checks Retargeting of a recorded performance; technical cleanup Stem separation, denoise, conformance Material Generated set extensions, objects, crowd agents, shots interleaved with photography Generated dialogue or scenes retained among authored material Partial synthesis of face, body or voice; synthetic background performers Generated cues or effects among authored sound Constituti ve Axis substantially generated (full-frame generation predominates) Story and dialogue substantially generated Synthetic performer(s) as principal characters Score or sound substantially generated AIFD-STD-DEF-001 / 5.1 AIFD-STD-DEF-001 9 Annex B (informative) — Worked examples B.1 A live-action feature with generated set extensions in 40 shots and a human script: image material, others none/assistive → AI-Hybrid. B.2 A fully generated short with human-written script and human voice acting: image constitutive only → AI- Native (image alone suffices). B.3 A documentary using AI solely for archive restoration and subtitling: no axis above assistive → AI- Assisted. B.4 An animation whose score is AI-composed but all visuals hand-animated: sound material → AI-Hybrid. B.5 A drama starring a digital replica of a deceased actor, otherwise conventional: performance constitutive → not AI-Native (one non-image axis); → AI-Hybrid, with AIFD-STD-RGT-001 clause 5 engaged. B.6 An anthology of five shorts, one AI-Native: parts classified individually; the anthology as a whole labels AI-Native (contains) per 5.5. B.7 A re-release with AI colourisation of a monochrome classic: image material in the new version → the re-release is AI-Hybrid; the original's classification is unaffected. B.8 A work whose licensed stock footage cannot be traced: image axis undetermined; least-favourable rule applies until traced. Annex R Revision history EDITIO N ADOPTED PRINCIPAL CHANGES 1.0 2024-08-01 First edition. 2.0 2024-11-08 Four-axis matrix introduced; assistive/material/constitutive levels defined. 3.0 2025-02-21 Axis thresholds consolidated; fixed minima for synthetic performers. 3.1 2025-04-11 Editorial revision; terminology aligned with AIFD-STD-PRO-001. 4.0 2025-08-29 Digital replica defined; special cases (anthology, restoration, documentary, live) added. 5.0 2026-03-06 Undetermined-axis rule and least-favourable determination; machine-readable class records. 5.1 2026-06-30 Editorial revision; worked examples B.6–B.8 added. No change of requirements. AIFD-STD-DEF-001 / 5.1 AIFD-STD-DEF-001 10 Bibliography (informative) Regulation (EU) 2024/1689 (AI Act), Article 50 — transparency obligations for synthetic content, noting the Article 50(4) provision for artistic works. Measures for the Labelling of AI-Generated Synthetic Content (PRC, in force 2025-09-01). California SB 942, AI Transparency Act. C2PA Specification — Coalition for Content Provenance and Authenticity. Bibliographic entries situate this standard in its regulatory and technical landscape. They are informative only and create no requirement.