/**
 * Editing pipeline types.
 *
 * The editing pipeline is parallel to the generation pipeline. It consumes
 * existing video/audio files (either Venice-generated shots or raw footage),
 * transcribes them, produces an LLM-readable pack, then drives an EDL-based
 * cut with a self-evaluating QA loop.
 *
 * Design principle from browser-use/video-use: text is the primary surface;
 * pixels are consulted on demand via the timeline_view composite.
 */
export interface WordTiming {
    /** The word as transcribed (may include punctuation attached by whisper). */
    word: string;
    /** Seconds from the start of the source file. */
    startSec: number;
    endSec: number;
    /** Optional confidence 0..1 from the transcriber. */
    confidence?: number;
}
export interface AudioEvent {
    /**
     * Event label e.g. `(laughter)`, `(applause)`, `(sigh)`, `(music)`.
     * Whisper.cpp does not emit these natively; aligner can inject them from
     * script cues (`shot.sfx`) for generated content.
     */
    label: string;
    startSec: number;
    endSec: number;
}
export interface TakePhrase {
    /**
     * Speaker label. For generated content this is the character name from
     * `shot.characters[0]`. For real footage it defaults to `S0` (single-
     * speaker). Pyannote integration in a future phase will populate this.
     */
    speaker: string;
    startSec: number;
    endSec: number;
    text: string;
    words: WordTiming[];
}
export interface Take {
    /**
     * Short opaque id used in `takes_packed.md` headings. Pattern `CNNNN`
     * (e.g. `C0103`) to match video-use. Generated from the source file name
     * hash + ordinal.
     */
    id: string;
    /** Absolute path to the source media file. */
    file: string;
    /** Total duration in seconds (from ffprobe). */
    durationSec: number;
    phrases: TakePhrase[];
    audioEvents: AudioEvent[];
    /** ISO timestamp when this take was transcribed. */
    transcribedAt: string;
    /** Transcriber that produced this take (e.g. `whisper-cpp:base.en`). */
    transcriber: string;
    /** Mode: pure ASR vs forced-aligned ground-truth script. */
    mode: 'asr' | 'aligned';
}
export interface TakesPack {
    takes: Take[];
    /** ISO timestamp when the pack was assembled. */
    generatedAt: string;
    /** Short description of the source folder. */
    sourceLabel: string;
}
export interface TranscribeOptions {
    /** Language code or 'auto' for detection. */
    language?: string;
    /** Whisper model size: tiny, base, small, medium, large. */
    modelSize?: 'tiny' | 'base' | 'small' | 'medium' | 'large' | 'tiny.en' | 'base.en' | 'small.en' | 'medium.en';
    /** Word-level timestamp threshold (whisper.cpp `--word-thold`). */
    wordThreshold?: number;
    /** Max cpu threads for whisper.cpp. */
    threads?: number;
}
export interface TranscribeResult {
    /** Per-word timings for the entire source. */
    words: WordTiming[];
    /** Full text (joined words, preserving whisper's punctuation). */
    text: string;
    /** Language detected. */
    language: string;
    /** Duration of the source in seconds (from ffprobe, not the transcriber). */
    durationSec: number;
    /** Human-readable transcriber label, e.g. `whisper-cpp:base.en`. */
    transcriberLabel: string;
}
export interface TranscriberProvider {
    /** Unique id for this provider, used in Take.transcriber. */
    readonly id: string;
    /**
     * Transcribe a single audio or video file. Callers are responsible for
     * audio extraction; providers should accept any file ffmpeg can decode.
     */
    transcribe(file: string, options?: TranscribeOptions): Promise<TranscribeResult>;
    /**
     * Check that the provider's dependencies are available (binary on PATH,
     * API key present, etc.). Throws with a helpful install message if not.
     */
    assertAvailable(): Promise<void>;
}
export type EdlTransition = 'cut' | 'crossfade' | 'fade-to-black';
export interface EdlClip {
    /** Source take id (matches Take.id). */
    sourceId: string;
    /** Seconds into the source where this clip starts. */
    startSec: number;
    endSec: number;
    /** Trim extra frames off the head/tail (applied on top of start/end). */
    trimStartMs?: number;
    trimEndMs?: number;
    /** Transition entering this clip (from previous clip). Default 'cut'. */
    transitionIn?: EdlTransition;
    /** Transition leaving this clip (to next clip). Default 'cut'. */
    transitionOut?: EdlTransition;
    /** Transition duration when crossfade/fade-to-black. Default 250ms. */
    transitionMs?: number;
    /**
     * Optional human-readable note from the LLM explaining why this clip was
     * included. Preserved through re-renders for audit.
     */
    rationale?: string;
}
export interface EdlColorGrade {
    /** Named grade preset, e.g. `warm-cinematic`, `neutral-punch`. */
    preset?: string;
    /** Raw ffmpeg filter chain (e.g. `eq=contrast=1.05:saturation=1.1`). */
    filter?: string;
}
export interface Edl {
    clips: EdlClip[];
    /** Global audio fade at every cut boundary (video-use's "no pops" rule). */
    audioFadeMs: number;
    /** Optional global color grade applied to every clip. */
    colorGrade?: EdlColorGrade;
    /** Target container / codec settings. */
    output: {
        width?: number;
        height?: number;
        fps?: number;
        videoCodec: string;
        audioCodec: string;
        crf: number;
    };
}
export interface CutQaFinding {
    kind: 'visual-jump' | 'aspect-regression' | 'subtitle-overlap' | 'vo-truncation' | 'lighting-discontinuity' | 'audio-pop';
    /** Trailer-time seconds where the issue was observed. */
    atSec: number;
    /** Which EDL clip index(es) triggered the finding. */
    clipIndex?: number;
    severity: 'warn' | 'fail';
    message: string;
    /** Path to a timeline_view composite illustrating the issue. */
    evidencePath?: string;
}
export interface CutQaReport {
    iteration: number;
    generatedAt: string;
    findings: CutQaFinding[];
    passed: boolean;
}
export interface EditSession {
    project: string;
    sourcesDir: string;
    editDir: string;
    takesPackPath: string;
    sources: string[];
    edl: Edl;
    iterations: CutQaReport[];
    createdAt: string;
    updatedAt: string;
}
//# sourceMappingURL=types.d.ts.map