The six official documents are available in English, Traditional Chinese and Simplified Chinese. This page provides the English full text and PDFs. Check the document number, revision and language when citing a document.
HKAIIFF-FEST-2027-003 · v1.4 · en
Download English PDFPrinciples of Evaluation
The work comes first; evidence supports the assessment
Jurors first view or experience the complete work, then use necessary creative and process records to understand AI's role, human decisions and the claimed achievements. Process records support understanding; they do not replace the work's expression.
Three categories of evaluation subject
The Grand Prix assesses the film as a whole; the three Silver Awards assess identifiable creative methods, digital characters and world systems; the three Innovation Awards assess achievements in narrative, audiovisual expression and experience. The same material may support different observations, but the distinct contribution recognised by each award must be explained.
Eligibility and value are separate
AI-native eligibility, authority to represent the entry, necessary permissions and completeness of disclosure are checked first. Passing these checks does not automatically confer a higher artistic assessment; a high score cannot offset unresolved material eligibility or rights issues.
Form and tool neutrality
Jurors consider the work's form in relation to its creative intention and realised achievements. Fixed films, real-time generative works and interactive works have no predetermined hierarchy. Tool brands, generation counts, resolution, cost and open-source status do not automatically determine achievement.
Fair understanding and uncertainty
Jurors must consider language, cultural and medium contexts, avoiding the assumption that unfamiliar forms are deficient. Gaps in information, technical failures and creative choices must be recorded separately; unverified matters must not be inferred as fact. Where necessary, requests for evidence within a defined scope may be made.
Contribution is assessed on the basis of actual decision-making and work performed, not on fixed job classifications, company status or headcount. For collaborative works, relevant contributions must be identified; authorship or sole ownership rights must not be presumed from submission account, funding source or agency status.
Five-level scale and evidence
| Level | Meaning | Basis of judgement |
|---|---|---|
| Level 5: Exceptional | Achievement is clear and fully substantiated | Creative choices are precise; the work or output holds consistently in key contexts, forming an inspiring and distinctive expression. |
| Level 4: Outstanding | The achievement is clear and highly accomplished | Primary choices are effective; key claims are well-supported by material; local limitations do not undermine core achievement. |
| Level 3: Realised | Core objectives are achieved | The work or output is understandable and achieves primary objectives, but uniqueness, depth or stability remains limited. |
| Level 2: Limited | Core achievement is only partially realised | Important intent is not fully realised, or there is a marked gap between outcome, control and evidence. |
| Level 1: Not demonstrated | Core achievements are not substantiated | The work or output has not achieved the primary effect claimed, or existing evidence cannot support key assertions. |
Where evidence is not yet provided, materials are temporarily inaccessible, or a dimension is indeed inapplicable, record respectively as 'Pending Verification' or 'Not Applicable'; these may not be substituted with a zero score or minimum rating. Records must be updated upon submission of supplementary evidence.
Purpose of the evidence
The finished work and actual experience demonstrate the realised effect. Versions, alternative outputs, revisions and test records help identify creative choices. Source and permission records support eligibility and rights checks. These types of evidence are not interchangeable.
Award determination
Award reasons must identify the core achievement assessed as Outstanding or Exceptional and explain how it relates to the complete subject under consideration. The five levels do not constitute an automatic award threshold or score cut-off. Decisions follow the full viewing, deliberation and voting procedures in the Awards and Jury Regulations.
Evaluation of the Golden Unicorn Grand Prix
The subject is the complete AI-native film. Jurors should reach an overall judgement supported by the work itself: do its creative choices form an integrated expression that cannot be arbitrarily taken apart?
| Evaluation dimensions | Focus of observation |
|---|---|
| Expression and originality | Whether the work possesses a clear and distinctive perspective, and whether its choices transcend tool preset effects or common material combinations. |
| Emotional and intellectual power | How the work generates feeling, understanding, reflection, or cultural meaning; not limited to specific subject matter, emotional intensity, or narrative genre. |
| Unity of form and content | How materials, rules, or experiences involving AI serve the work; whether narrative, audiovisual, and medium choices support one another. |
| Overall accomplishment | Does the work hold together within its stated scope? For fixed films, assess the whole film. For dynamic works, assess representative experiences, the range of variation and whether the achievement holds over repeated experiences. |
Required materials
Complete works, creative intent, primary human contributions, and an explanation of AI involvement. For dynamic works, additional requirements include a stable version, suggested pathways, methods for independent juror exploration, and known limitations.
Jurors should answer
What is the strongest achievement? Where does it appear? How do the main choices work together? If the work has clear limitations, why does its overall achievement still hold, or how do those limitations weaken it?
Evaluation of the Silver Unicorn · Creative Method
The subject is a method demonstrated in a work and developed or substantially improved by the applicant team. It may be implemented by one person or involve multiple tools, agent systems and collaborating teams.
| Evaluation dimensions | Focus of observation |
|---|---|
| Problem and Creative Value | What specific creative problem does the method solve, and how is its value observed within the work. |
| Intent, Choice, and Control | How are goals set, alternative outputs selected, deviations checked, results revised and necessary consistency maintained? |
| Validation and Robustness | Can key claims be verified? Can the goal be reproducibly achieved under stated conditions, and are failure conditions understood? |
| Sustained Use and Handover | Can the method be applied to new shots or contexts? Are dependencies, human input, resource consumption, and documentation clear? |
Required Evidence
At least one eligible entry from this edition, an explanation of the method, input-output comparisons, key human decisions and one representative repeat test. Claims of improved speed, cost or quality require comparable baselines and conditions.
Verification Method
Restricted test environments, controlled demonstrations, step-by-step explanations with secrets removed, or other means sufficient to verify claims may be used. Closed-source submissions are not automatically penalised, but achievements that cannot be verified at all cannot be recognised on mere assertion.
Evaluation of the Silver Unicorn · Digital Character
The subject of assessment is a digital character with a recognisable identity and performance in the work. The character may be human, non-human, an abstract entity, or another form that functions as a character.
| Evaluation dimensions | Focus of observation |
|---|---|
| Identity and recognisability | Whether appearance, voice, behaviour, or other signifiers collectively constitute a distinct identity, rather than relying solely on novel design. |
| Behaviour and emotional expression | Whether motivation, reaction, performance, and change are intelligible and generate expressive power. |
| Cross-situational continuity | Whether the character can maintain a core identity across different situations while presenting grounded changes. |
| Creative control and extension | Whether the team can consciously adjust the performance and articulate constraints; whether the character supports subsequent creation rather than existing only within a single accidental fragment. |
Required Evidence
The character definition, realised performances in at least two different situations, locators in the associated work, and the principal creative, performance and revision contributions. Both situations may occur in the same film.
Verification Method
Fixed characters are assessed against actual segments, versions, or revision evidence; interactive characters are assessed against their declared scope of response. Unpromised chat capabilities must not be required for fixed film characters, nor should individual successful dialogues be treated as proof of total capability.
Assessment of the Worldbuilding Award
The subject is a world system realised in a work, with coherent rules and relationships. Its value depends on how it is established and its capacity to support narrative.
| Evaluation dimensions | Focus of observation |
|---|---|
| Uniqueness and overall coherence | Whether space, culture, audiovisual style, and other elements form a recognisable and mutually supportive system. |
| Rules, relationships, and causality | Whether characters and events are influenced by world rules; whether behaviour yields intelligible consequences. |
| Capacity to support stories and experiences | Whether the world can support characters, conflict, emotion, and multiple realised situations. |
| Extension and Maintenance Capability | How core rules are sustained when adding characters, scenes, or events; whether versioning, dependency, and constraints are clearly defined. |
Required Evidence
World overview, primary rules, at least two realised scenarios, and their correspondence to the submitted work. Claims of extension capability must be supported by new scenarios, actual testing, or concrete verifiable implementation materials.
Verification Method
Jurors first observe the rules in the work, then consult supplementary world descriptions. Such descriptions explain the world; they cannot replace a world never realised in the work. Real-time worlds also require checks of state, memory and recovery. Worlds in fixed films are assessed through their realised expression.
Evaluation of the Unicorn · Narrative Innovation Award
The subject is the work's achievement in organising story, information and meaning. It may present a clear story or use essayistic, poetic, non-narrative or other forms that organise meaning.
| Evaluation dimensions | Focus of observation |
|---|---|
| Narrative Intent and Perspective | How the work establishes a position for viewing or understanding, and whether the formal choice serves a clear purpose. |
| Time, Information, and Causality | How narrative materials are arranged, concealed, repeated, or altered; whether the structure generates necessary understanding and tension. |
| Character, Emotion, and Meaning | How the narrative arrangement forms character experience, emotional processes, thought, or cultural meaning. |
| Change and Overall Coherence | Whether parts within a fixed structure support one another; whether different paths in a dynamic narrative maintain the core logic they have promised. |
Required Evidence
Complete works and timecodes, segments, or paths that explain key arrangements. For dynamic works, provide representative paths, primary ranges of variation, and generative constraints; traditional screenplay format is not required for all submissions.
Boundaries with Other Awards
The Narrative Award focuses on how meaning is organised; the Experiential Award focuses on audience engagement and the work's response. An interactive arrangement may affect both; juries must separately explain the narrative effect and the participatory effect.
Evaluation of the Unicorn · Audiovisual Innovation Award
The subject is the integrated language formed by image, sound, movement, space and rhythm. Deliberately silent or static works are assessed in relation to their expressive purpose.
| Evaluation dimensions | Focus of observation |
|---|---|
| Form and Perception | Whether the audiovisual choices create a unique, perceptible, and work-relevant experience. |
| Relationships between elements | How sound alters the understanding of imagery, how motion and editing organise time, and how space and rhythm act together. |
| Intent and Control | Whether the creator consciously selects, revises, and maintains the required language; how accidental results are incorporated into the whole. |
| Continuity of the overall language | Whether key expressions hold throughout the work, and whether local deviations are meaningful or undermine the experience. |
Required Evidence
Complete work, key segments, and concise creative statements. Where necessary, verification of pre- and post-revision versions, audio-visual relationships, or editing choices may be required; submission of a commercial-grade multi-track master is not mandatory for all entrants.
Boundaries with Other Awards
The audiovisual award concerns the expressive effect realised in the work; the creative method award concerns a verifiable method capable of continued application. High technical specifications, fast generation or an elaborate workflow alone do not justify an audiovisual award.
Assessment of the Experiential Innovation Award
The subject is the deliberate design of viewing paths, participation, responses from the work or shared experience. Applicants must identify the specific change in the relationship with the audience.
| Evaluation dimensions | Focus of observation |
|---|---|
| Understandability of participation | Whether the viewer can understand how to enter, choose, act, or exit; and whether the interface is consistent with the work. |
| Action and Consequence | Whether the viewer's actions bring meaningful change and relate to the work's content or affect. |
| Response and Creative Intent | Whether the work maintains clear boundaries, expression, and necessary continuity when responding to the viewer. |
| Completeness and different paths | Whether the experience holds within the committed scope under different participation modes, repeated entry, or multi-user scenarios. |
Required Evidence
Provide a playable or experienceable version, operating and exit instructions, suggested duration, representative paths, known limitations and an explanation of audience-data use. If audience-test results are cited, describe the test conditions; sample size cannot replace professional judgement.
Verification Method
Jurors must complete at least one creator-recommended path and one path of their own choosing, with appropriate repeat tests where material randomness or response differences exist. Testing must remain within the declared scope of use. Unauthorised destructive testing is prohibited.
Viewing and Verification Procedures
| Procedures | Public Method |
|---|---|
| Independent Review | Complete the viewing or experience before consulting necessary creative evidence. Jurors first record their own observations independently; other people's comments cannot replace those observations. |
| Version Confirmation | Record version date, file or build identifier, primary tools and runtime conditions; all jurors must use a comparable submission version. |
| Dynamic Testing | Complete the recommended path, a juror-chosen path and necessary repeat tests. Record reset procedures, retained state and material differences. |
| Troubleshooting | Distinguish between limitations inherent to the work, third-party service interruptions, and juror environment issues; provide consistent opportunities for repair or re-testing. |
| Limited Supplementary Evidence | Requests must identify the specific claim requiring clarification. They must not demand trade secrets unrelated to the assessment or model-training information the applicant does not possess. |
| Deliberation | Discuss specific observations, differences in evidence and creative achievements. Apply the recusal, quorum and voting procedures in the Awards and Jury Regulations. |
Dynamic works need not generate identical content on every run, but applicants must explain the core conditions that should remain consistent. Jurors check whether those conditions hold. They must not treat every variation as failure, or an isolated success as stable capability.
If a work depends on a discontinued model or service, the applicant should propose verifiable archives, a controlled alternative environment or a demonstration. If these still cannot verify a particular achievement, the festival should explain the effect on that application without automatically disqualifying a film that can be viewed in full.
Evaluation records and public explanation
| Record items | Content to be recorded |
|---|---|
| Subject and scope | The work or achievement title, award, version, and actual scope viewed or tested. |
| Specific observations | Key timestamps, scenes, paths, or test conditions, together with the actual observations recorded. |
| Dimensional evaluation | Record the five-level judgement, corresponding rationale, items pending verification, and items not applicable. |
| Limitations and uncertainties | Insufficient materials, technical issues, cultural or linguistic comprehension limitations, and follow-up verification results. |
| Overall recommendation | Core achievement, primary limitations, and whether the work is supported for progression to the next stage; do not substitute reasoned judgement with an arithmetic total score. |
| Procedural confirmation | Completion of review, conflict-of-interest declaration, necessary recusal, and voting eligibility. |
Public jury explanation
Published award reasons should enable audiences to understand the specific achievement recognised, how it is realised in the work or achievement, and why it relates to the award. Cite actual examples suitable for publication without disclosing unauthorised private logs, complete prompts, confidential methods or personal data.
Responsible commentary
Do not substitute model brands, budgets, nationality, creator reputation or publicity for observation. Procedural checks must not be presented as legal determinations of rights, and unrealised capabilities must not be described as demonstrated achievements.
Procedural Safeguards
Jurors make artistic judgements through the procedures in the Awards and Jury Regulations. Procedural review may be requested under those Regulations regarding eligibility facts, conflicts of interest or material procedural issues. Enquiries: info@hkaiiff.org.