sonilo 0.2.1 → 0.4.0

package/dist/index.d.cts

@@ -51,6 +51,10 @@ interface TextToMusicParams {
     prompt: string;
     duration: number;
     segments?: Segment[];
+    /** "stream" (default) or "async" (required by `submit()` and `output_format: "wav"`). */
+    mode?: "stream" | "async";
+    /** Container for the async result. `wav` requires `mode: "async"`. Defaults to m4a server-side. */
+    outputFormat?: "m4a" | "wav";
     /** Bounds the stream: aborting this cancels the in-flight generation.
      * Passed straight through to `fetch` — it is never rewrapped as
      * RequestTimeoutError, since the client's own absolute timeout does not
@@ -67,8 +71,28 @@ interface VideoToMusicParams {
     /** Bounds the stream: aborting this cancels the in-flight generation.
      * Passed straight through to `fetch` — it is never rewrapped as
      * RequestTimeoutError, since the client's own absolute timeout does not
-     * apply to streaming music generation. */
+     * apply to streaming music generation. Only meaningful for `stream()`/
+     * `generate()`; `submit()` ignores it. */
     signal?: AbortSignal;
+    /** "stream" (the default, used by `stream()`/`generate()`) or "async"
+     * (required for `submit()`, and for `isolateVocals`). Only consulted by
+     * `submit()` — `stream()`/`generate()` always request a stream. */
+    mode?: "stream" | "async";
+    /** Split the generated track into a vocals-only stem alongside the mix.
+     * Requires `mode: "async"`; if `mode` is left unset it defaults to
+     * "async" automatically. Only usable via `submit()` — the backend
+     * rejects it on the plain stream. */
+    isolateVocals?: boolean;
+    /** Keep the source speech/vocals in the async result. Current name for
+     * `isolateVocals`; both are accepted and OR'd server-side. Requires
+     * `mode: "async"` (auto-selected by `submit()`). */
+    preserveSpeech?: boolean;
+    /** Container for the async result. `wav` requires async. Defaults to m4a. */
+    outputFormat?: "m4a" | "wav";
+    /** Duck the generated music under the source voice at finalize time.
+     * Default-ON server-side in async mode: leave unset to keep it on, pass
+     * `false` to opt out. Free, best-effort; only valid on `submit()`. */
+    ducking?: boolean;
 }
 interface AccountServices {
     available_services: string[];
@@ -120,20 +144,29 @@ interface SfxError {
     code?: string;
     message?: string;
 }
-/** State of an SFX task (`tasks.get`) or its final result (`wait`/`generate`). */
-interface SfxResult {
+/**
+ * Common shape of any polled task (`tasks.get`/`tasks.wait`), regardless of
+ * which endpoint created it. `Tasks.get`/`Tasks.wait` are generic over this so
+ * each endpoint's result type (e.g. `SfxResult`, `MusicTaskResult`) can add
+ * its own `audio`/media fields while sharing the status/error/refund
+ * bookkeeping the poller relies on.
+ */
+interface BaseTaskResult {
     task_id: string;
     type?: string;
     status: "processing" | "succeeded" | "failed" | (string & {});
-    audio?: SfxMedia;
-    /** Kept for backward compatibility; no longer populated — video-to-sfx returns audio only. */
-    video?: SfxMedia;
     /** Only present when the account's task-field whitelist enables cost. */
     cost?: number;
     error?: SfxError;
     refunded?: boolean;
     [key: string]: unknown;
 }
+/** State of an SFX task (`tasks.get`) or its final result (`wait`/`generate`). */
+interface SfxResult extends BaseTaskResult {
+    audio?: SfxMedia;
+    /** Kept for backward compatibility; no longer populated — video-to-sfx returns audio only. */
+    video?: SfxMedia;
+}
 interface TextToSfxParams {
     prompt: string;
     duration: number;
@@ -146,12 +179,66 @@ interface VideoToSfxParams {
     segments?: SfxSegment[];
     audioFormat?: SfxAudioFormat;
 }
+/** One decoded audio stream of an async video-to-music result. Unlike SFX,
+ * `audio` on a music task is always an array — even without `isolateVocals` —
+ * since a music generation can carry more than one output stream. */
+interface MusicMediaEntry extends SfxMedia {
+    stream_index: number;
+    sample_rate?: number;
+    channels?: number;
+}
+/** One muxed audio+video-aligned output, present only when `isolateVocals`
+ * is set. */
+interface MusicMuxEntry extends SfxMedia {
+    stream_index: number;
+}
+interface MusicTitle {
+    title: string;
+    summary?: string;
+    display_tags?: string[];
+}
+/** State of an async video-to-music task (`tasks.get`) or its final result
+ * (`tasks.wait<MusicTaskResult>()`). Only reachable via `videoToMusic.submit()`
+ * with `mode: "async"`. */
+interface MusicTaskResult extends BaseTaskResult {
+    audio?: MusicMediaEntry[];
+    /** Vocals-only stem; present only when `isolateVocals` was requested. */
+    vocals?: SfxMedia;
+    /** Muxed output per stream; present only when `isolateVocals` was requested. */
+    mux?: MusicMuxEntry[];
+    /** Music ducked under the source voice; present only when `ducking` ran. */
+    ducked?: MusicMediaEntry[];
+    title?: MusicTitle;
+    duration_seconds?: number;
+}
 interface WaitOptions {
     /** Milliseconds between polls. Default 2000. */
     pollInterval?: number;
     /** Overall deadline in milliseconds. Default 600000. */
     timeout?: number;
 }
+/** Result of an async video-to-video task (`videoToVideoMusic`/`videoToVideoSfx`):
+ * a re-hosted video with generated music or SFX muxed in. */
+interface VideoResult extends BaseTaskResult {
+    video?: SfxMedia;
+    duration_seconds?: number;
+}
+interface VideoToVideoMusicParams {
+    video?: VideoInput;
+    videoUrl?: string;
+    prompt?: string;
+    /** Keep the source speech/vocals in the output. Both this and the legacy
+     * `isolateVocals` are accepted and OR'd server-side. */
+    preserveSpeech?: boolean;
+    /** @deprecated Legacy alias for `preserveSpeech`. */
+    isolateVocals?: boolean;
+}
+interface VideoToVideoSfxParams {
+    video?: VideoInput;
+    videoUrl?: string;
+    prompt?: string;
+    segments?: SfxSegment[];
+}
 
 declare class Account {
     private readonly client;
@@ -165,10 +252,22 @@ declare class Account {
 declare class Tasks {
     private readonly client;
     constructor(client: SoniloClient);
-    /** Fetch current task state. Never throws on a failed status. */
-    get(taskId: string): Promise<SfxResult>;
-    /** Poll until the task is terminal; throw on failure or deadline. */
-    wait(taskId: string, opts?: WaitOptions): Promise<SfxResult>;
+    /**
+     * Fetch current task state. Never throws on a failed status.
+     *
+     * Generic over the result shape so callers can request the endpoint-
+     * specific type, e.g. `client.tasks.get<MusicTaskResult>(taskId)`.
+     * Defaults to `SfxResult` for back-compat.
+     */
+    get<T extends BaseTaskResult = SfxResult>(taskId: string): Promise<T>;
+    /**
+     * Poll until the task is terminal; throw on failure or deadline.
+     *
+     * Generic over the result shape, e.g.
+     * `client.tasks.wait<MusicTaskResult>(taskId)`. Defaults to `SfxResult`
+     * for back-compat.
+     */
+    wait<T extends BaseTaskResult = SfxResult>(taskId: string, opts?: WaitOptions): Promise<T>;
 }
 
 declare class TextToMusic {
@@ -178,6 +277,12 @@ declare class TextToMusic {
     stream(params: TextToMusicParams): AsyncGenerator<StreamEvent, void, undefined>;
     /** Generate and buffer the whole track; throws GenerationError on stream errors. */
     generate(params: TextToMusicParams): Promise<Track>;
+    /**
+     * Submit an async text-to-music task; poll with
+     * `client.tasks.wait<MusicTaskResult>(task.task_id)`. Required for
+     * `outputFormat: "wav"`. `stream()`/`generate()` remain the streaming path.
+     */
+    submit(params: TextToMusicParams): Promise<SfxTask>;
 }
 
 declare class VideoToMusic {
@@ -185,6 +290,13 @@ declare class VideoToMusic {
     constructor(client: SoniloClient);
     stream(params: VideoToMusicParams): AsyncGenerator<StreamEvent, void, undefined>;
     generate(params: VideoToMusicParams): Promise<Track>;
+    /**
+     * Submit an async video-to-music task; poll its result with
+     * `client.tasks.wait<MusicTaskResult>(task.task_id)`. Required for
+     * `isolateVocals` — the backend rejects vocal isolation on the plain
+     * stream, and it only ever runs in async mode.
+     */
+    submit(params: VideoToMusicParams): Promise<SfxTask>;
 }
 
 declare class TextToSfx {
@@ -201,6 +313,25 @@ declare class VideoToSfx {
     generate(params: VideoToSfxParams, opts?: WaitOptions): Promise<SfxResult>;
 }
 
+/** Generate an original score for a video and get back a re-hosted video with
+ * the music muxed in. Async only: `submit()` returns a task ack; poll with
+ * `client.tasks.wait<VideoResult>(id)`, or use `generate()` to do both. */
+declare class VideoToVideoMusic {
+    private readonly client;
+    constructor(client: SoniloClient);
+    submit(params: VideoToVideoMusicParams): Promise<SfxTask>;
+    generate(params: VideoToVideoMusicParams, opts?: WaitOptions): Promise<VideoResult>;
+}
+
+/** Generate sound effects for a video and get back a re-hosted video with the
+ * SFX muxed in. Async only. */
+declare class VideoToVideoSfx {
+    private readonly client;
+    constructor(client: SoniloClient);
+    submit(params: VideoToVideoSfxParams): Promise<SfxTask>;
+    generate(params: VideoToVideoSfxParams, opts?: WaitOptions): Promise<VideoResult>;
+}
+
 interface SoniloClientOptions {
     /** Defaults to the SONILO_API_KEY environment variable (Node.js only). */
     apiKey?: string;
@@ -224,6 +355,8 @@ declare class SoniloClient {
     readonly videoToMusic: VideoToMusic;
     readonly textToSfx: TextToSfx;
     readonly videoToSfx: VideoToSfx;
+    readonly videoToVideoMusic: VideoToVideoMusic;
+    readonly videoToVideoSfx: VideoToVideoSfx;
     constructor(options?: SoniloClientOptions);
     /**
      * Perform an authenticated request; throws a typed error on non-2xx.
@@ -292,6 +425,6 @@ declare class RequestTimeoutError extends SoniloError {
 /** Fetch a result media file. The URL is presigned — no API key is sent. */
 declare function download(media: SfxMedia | undefined, fetchFn?: typeof globalThis.fetch, timeout?: number): Promise<Uint8Array>;
 
-declare const VERSION = "0.2.0";
+declare const VERSION = "0.4.0";
 
-export { APIError, type AccountServices, type AudioChunkEvent, AuthenticationError, BadRequestError, type CompleteEvent, type CostEvent, type CostInfo, DEFAULT_TIMEOUT_MS, type DailyUsage, type ErrorEvent, GenerationError, PaymentRequiredError, RateLimitError, RequestTimeoutError, type Segment, type SegmentLabel, type SfxAudioFormat, type SfxError, type SfxMedia, type SfxResult, type SfxSegment, type SfxTask, SoniloClient, type SoniloClientOptions, SoniloError, type StreamEvent, TaskFailedError, TaskTimeoutError, type TextToMusicParams, type TextToSfxParams, type TitleEvent, type Track, type UnknownEvent, type UsageResponse, type UsageSummary, VERSION, type VideoInput, type VideoToMusicParams, type VideoToSfxParams, type WaitOptions, download, isAudioChunkEvent, isErrorEvent };
+export { APIError, type AccountServices, type AudioChunkEvent, AuthenticationError, BadRequestError, type BaseTaskResult, type CompleteEvent, type CostEvent, type CostInfo, DEFAULT_TIMEOUT_MS, type DailyUsage, type ErrorEvent, GenerationError, type MusicMediaEntry, type MusicMuxEntry, type MusicTaskResult, type MusicTitle, PaymentRequiredError, RateLimitError, RequestTimeoutError, type Segment, type SegmentLabel, type SfxAudioFormat, type SfxError, type SfxMedia, type SfxResult, type SfxSegment, type SfxTask, SoniloClient, type SoniloClientOptions, SoniloError, type StreamEvent, TaskFailedError, TaskTimeoutError, type TextToMusicParams, type TextToSfxParams, type TitleEvent, type Track, type UnknownEvent, type UsageResponse, type UsageSummary, VERSION, type VideoInput, type VideoResult, type VideoToMusicParams, type VideoToSfxParams, type VideoToVideoMusicParams, type VideoToVideoSfxParams, type WaitOptions, download, isAudioChunkEvent, isErrorEvent };

package/dist/index.d.ts

@@ -51,6 +51,10 @@ interface TextToMusicParams {
     prompt: string;
     duration: number;
     segments?: Segment[];
+    /** "stream" (default) or "async" (required by `submit()` and `output_format: "wav"`). */
+    mode?: "stream" | "async";
+    /** Container for the async result. `wav` requires `mode: "async"`. Defaults to m4a server-side. */
+    outputFormat?: "m4a" | "wav";
     /** Bounds the stream: aborting this cancels the in-flight generation.
      * Passed straight through to `fetch` — it is never rewrapped as
      * RequestTimeoutError, since the client's own absolute timeout does not
@@ -67,8 +71,28 @@ interface VideoToMusicParams {
     /** Bounds the stream: aborting this cancels the in-flight generation.
      * Passed straight through to `fetch` — it is never rewrapped as
      * RequestTimeoutError, since the client's own absolute timeout does not
-     * apply to streaming music generation. */
+     * apply to streaming music generation. Only meaningful for `stream()`/
+     * `generate()`; `submit()` ignores it. */
     signal?: AbortSignal;
+    /** "stream" (the default, used by `stream()`/`generate()`) or "async"
+     * (required for `submit()`, and for `isolateVocals`). Only consulted by
+     * `submit()` — `stream()`/`generate()` always request a stream. */
+    mode?: "stream" | "async";
+    /** Split the generated track into a vocals-only stem alongside the mix.
+     * Requires `mode: "async"`; if `mode` is left unset it defaults to
+     * "async" automatically. Only usable via `submit()` — the backend
+     * rejects it on the plain stream. */
+    isolateVocals?: boolean;
+    /** Keep the source speech/vocals in the async result. Current name for
+     * `isolateVocals`; both are accepted and OR'd server-side. Requires
+     * `mode: "async"` (auto-selected by `submit()`). */
+    preserveSpeech?: boolean;
+    /** Container for the async result. `wav` requires async. Defaults to m4a. */
+    outputFormat?: "m4a" | "wav";
+    /** Duck the generated music under the source voice at finalize time.
+     * Default-ON server-side in async mode: leave unset to keep it on, pass
+     * `false` to opt out. Free, best-effort; only valid on `submit()`. */
+    ducking?: boolean;
 }
 interface AccountServices {
     available_services: string[];
@@ -120,20 +144,29 @@ interface SfxError {
     code?: string;
     message?: string;
 }
-/** State of an SFX task (`tasks.get`) or its final result (`wait`/`generate`). */
-interface SfxResult {
+/**
+ * Common shape of any polled task (`tasks.get`/`tasks.wait`), regardless of
+ * which endpoint created it. `Tasks.get`/`Tasks.wait` are generic over this so
+ * each endpoint's result type (e.g. `SfxResult`, `MusicTaskResult`) can add
+ * its own `audio`/media fields while sharing the status/error/refund
+ * bookkeeping the poller relies on.
+ */
+interface BaseTaskResult {
     task_id: string;
     type?: string;
     status: "processing" | "succeeded" | "failed" | (string & {});
-    audio?: SfxMedia;
-    /** Kept for backward compatibility; no longer populated — video-to-sfx returns audio only. */
-    video?: SfxMedia;
     /** Only present when the account's task-field whitelist enables cost. */
     cost?: number;
     error?: SfxError;
     refunded?: boolean;
     [key: string]: unknown;
 }
+/** State of an SFX task (`tasks.get`) or its final result (`wait`/`generate`). */
+interface SfxResult extends BaseTaskResult {
+    audio?: SfxMedia;
+    /** Kept for backward compatibility; no longer populated — video-to-sfx returns audio only. */
+    video?: SfxMedia;
+}
 interface TextToSfxParams {
     prompt: string;
     duration: number;
@@ -146,12 +179,66 @@ interface VideoToSfxParams {
     segments?: SfxSegment[];
     audioFormat?: SfxAudioFormat;
 }
+/** One decoded audio stream of an async video-to-music result. Unlike SFX,
+ * `audio` on a music task is always an array — even without `isolateVocals` —
+ * since a music generation can carry more than one output stream. */
+interface MusicMediaEntry extends SfxMedia {
+    stream_index: number;
+    sample_rate?: number;
+    channels?: number;
+}
+/** One muxed audio+video-aligned output, present only when `isolateVocals`
+ * is set. */
+interface MusicMuxEntry extends SfxMedia {
+    stream_index: number;
+}
+interface MusicTitle {
+    title: string;
+    summary?: string;
+    display_tags?: string[];
+}
+/** State of an async video-to-music task (`tasks.get`) or its final result
+ * (`tasks.wait<MusicTaskResult>()`). Only reachable via `videoToMusic.submit()`
+ * with `mode: "async"`. */
+interface MusicTaskResult extends BaseTaskResult {
+    audio?: MusicMediaEntry[];
+    /** Vocals-only stem; present only when `isolateVocals` was requested. */
+    vocals?: SfxMedia;
+    /** Muxed output per stream; present only when `isolateVocals` was requested. */
+    mux?: MusicMuxEntry[];
+    /** Music ducked under the source voice; present only when `ducking` ran. */
+    ducked?: MusicMediaEntry[];
+    title?: MusicTitle;
+    duration_seconds?: number;
+}
 interface WaitOptions {
     /** Milliseconds between polls. Default 2000. */
     pollInterval?: number;
     /** Overall deadline in milliseconds. Default 600000. */
     timeout?: number;
 }
+/** Result of an async video-to-video task (`videoToVideoMusic`/`videoToVideoSfx`):
+ * a re-hosted video with generated music or SFX muxed in. */
+interface VideoResult extends BaseTaskResult {
+    video?: SfxMedia;
+    duration_seconds?: number;
+}
+interface VideoToVideoMusicParams {
+    video?: VideoInput;
+    videoUrl?: string;
+    prompt?: string;
+    /** Keep the source speech/vocals in the output. Both this and the legacy
+     * `isolateVocals` are accepted and OR'd server-side. */
+    preserveSpeech?: boolean;
+    /** @deprecated Legacy alias for `preserveSpeech`. */
+    isolateVocals?: boolean;
+}
+interface VideoToVideoSfxParams {
+    video?: VideoInput;
+    videoUrl?: string;
+    prompt?: string;
+    segments?: SfxSegment[];
+}
 
 declare class Account {
     private readonly client;
@@ -165,10 +252,22 @@ declare class Account {
 declare class Tasks {
     private readonly client;
     constructor(client: SoniloClient);
-    /** Fetch current task state. Never throws on a failed status. */
-    get(taskId: string): Promise<SfxResult>;
-    /** Poll until the task is terminal; throw on failure or deadline. */
-    wait(taskId: string, opts?: WaitOptions): Promise<SfxResult>;
+    /**
+     * Fetch current task state. Never throws on a failed status.
+     *
+     * Generic over the result shape so callers can request the endpoint-
+     * specific type, e.g. `client.tasks.get<MusicTaskResult>(taskId)`.
+     * Defaults to `SfxResult` for back-compat.
+     */
+    get<T extends BaseTaskResult = SfxResult>(taskId: string): Promise<T>;
+    /**
+     * Poll until the task is terminal; throw on failure or deadline.
+     *
+     * Generic over the result shape, e.g.
+     * `client.tasks.wait<MusicTaskResult>(taskId)`. Defaults to `SfxResult`
+     * for back-compat.
+     */
+    wait<T extends BaseTaskResult = SfxResult>(taskId: string, opts?: WaitOptions): Promise<T>;
 }
 
 declare class TextToMusic {
@@ -178,6 +277,12 @@ declare class TextToMusic {
     stream(params: TextToMusicParams): AsyncGenerator<StreamEvent, void, undefined>;
     /** Generate and buffer the whole track; throws GenerationError on stream errors. */
     generate(params: TextToMusicParams): Promise<Track>;
+    /**
+     * Submit an async text-to-music task; poll with
+     * `client.tasks.wait<MusicTaskResult>(task.task_id)`. Required for
+     * `outputFormat: "wav"`. `stream()`/`generate()` remain the streaming path.
+     */
+    submit(params: TextToMusicParams): Promise<SfxTask>;
 }
 
 declare class VideoToMusic {
@@ -185,6 +290,13 @@ declare class VideoToMusic {
     constructor(client: SoniloClient);
     stream(params: VideoToMusicParams): AsyncGenerator<StreamEvent, void, undefined>;
     generate(params: VideoToMusicParams): Promise<Track>;
+    /**
+     * Submit an async video-to-music task; poll its result with
+     * `client.tasks.wait<MusicTaskResult>(task.task_id)`. Required for
+     * `isolateVocals` — the backend rejects vocal isolation on the plain
+     * stream, and it only ever runs in async mode.
+     */
+    submit(params: VideoToMusicParams): Promise<SfxTask>;
 }
 
 declare class TextToSfx {
@@ -201,6 +313,25 @@ declare class VideoToSfx {
     generate(params: VideoToSfxParams, opts?: WaitOptions): Promise<SfxResult>;
 }
 
+/** Generate an original score for a video and get back a re-hosted video with
+ * the music muxed in. Async only: `submit()` returns a task ack; poll with
+ * `client.tasks.wait<VideoResult>(id)`, or use `generate()` to do both. */
+declare class VideoToVideoMusic {
+    private readonly client;
+    constructor(client: SoniloClient);
+    submit(params: VideoToVideoMusicParams): Promise<SfxTask>;
+    generate(params: VideoToVideoMusicParams, opts?: WaitOptions): Promise<VideoResult>;
+}
+
+/** Generate sound effects for a video and get back a re-hosted video with the
+ * SFX muxed in. Async only. */
+declare class VideoToVideoSfx {
+    private readonly client;
+    constructor(client: SoniloClient);
+    submit(params: VideoToVideoSfxParams): Promise<SfxTask>;
+    generate(params: VideoToVideoSfxParams, opts?: WaitOptions): Promise<VideoResult>;
+}
+
 interface SoniloClientOptions {
     /** Defaults to the SONILO_API_KEY environment variable (Node.js only). */
     apiKey?: string;
@@ -224,6 +355,8 @@ declare class SoniloClient {
     readonly videoToMusic: VideoToMusic;
     readonly textToSfx: TextToSfx;
     readonly videoToSfx: VideoToSfx;
+    readonly videoToVideoMusic: VideoToVideoMusic;
+    readonly videoToVideoSfx: VideoToVideoSfx;
     constructor(options?: SoniloClientOptions);
     /**
      * Perform an authenticated request; throws a typed error on non-2xx.
@@ -292,6 +425,6 @@ declare class RequestTimeoutError extends SoniloError {
 /** Fetch a result media file. The URL is presigned — no API key is sent. */
 declare function download(media: SfxMedia | undefined, fetchFn?: typeof globalThis.fetch, timeout?: number): Promise<Uint8Array>;
 
-declare const VERSION = "0.2.0";
+declare const VERSION = "0.4.0";
 
-export { APIError, type AccountServices, type AudioChunkEvent, AuthenticationError, BadRequestError, type CompleteEvent, type CostEvent, type CostInfo, DEFAULT_TIMEOUT_MS, type DailyUsage, type ErrorEvent, GenerationError, PaymentRequiredError, RateLimitError, RequestTimeoutError, type Segment, type SegmentLabel, type SfxAudioFormat, type SfxError, type SfxMedia, type SfxResult, type SfxSegment, type SfxTask, SoniloClient, type SoniloClientOptions, SoniloError, type StreamEvent, TaskFailedError, TaskTimeoutError, type TextToMusicParams, type TextToSfxParams, type TitleEvent, type Track, type UnknownEvent, type UsageResponse, type UsageSummary, VERSION, type VideoInput, type VideoToMusicParams, type VideoToSfxParams, type WaitOptions, download, isAudioChunkEvent, isErrorEvent };
+export { APIError, type AccountServices, type AudioChunkEvent, AuthenticationError, BadRequestError, type BaseTaskResult, type CompleteEvent, type CostEvent, type CostInfo, DEFAULT_TIMEOUT_MS, type DailyUsage, type ErrorEvent, GenerationError, type MusicMediaEntry, type MusicMuxEntry, type MusicTaskResult, type MusicTitle, PaymentRequiredError, RateLimitError, RequestTimeoutError, type Segment, type SegmentLabel, type SfxAudioFormat, type SfxError, type SfxMedia, type SfxResult, type SfxSegment, type SfxTask, SoniloClient, type SoniloClientOptions, SoniloError, type StreamEvent, TaskFailedError, TaskTimeoutError, type TextToMusicParams, type TextToSfxParams, type TitleEvent, type Track, type UnknownEvent, type UsageResponse, type UsageSummary, VERSION, type VideoInput, type VideoResult, type VideoToMusicParams, type VideoToSfxParams, type VideoToVideoMusicParams, type VideoToVideoSfxParams, type WaitOptions, download, isAudioChunkEvent, isErrorEvent };

package/dist/index.js

@@ -136,12 +136,24 @@ var Tasks = class {
   constructor(client) {
     this.client = client;
   }
-  /** Fetch current task state. Never throws on a failed status. */
+  /**
+   * Fetch current task state. Never throws on a failed status.
+   *
+   * Generic over the result shape so callers can request the endpoint-
+   * specific type, e.g. `client.tasks.get<MusicTaskResult>(taskId)`.
+   * Defaults to `SfxResult` for back-compat.
+   */
   async get(taskId) {
     const res = await this.client.request(`/v1/tasks/${encodeURIComponent(taskId)}`);
     return await res.json();
   }
-  /** Poll until the task is terminal; throw on failure or deadline. */
+  /**
+   * Poll until the task is terminal; throw on failure or deadline.
+   *
+   * Generic over the result shape, e.g.
+   * `client.tasks.wait<MusicTaskResult>(taskId)`. Defaults to `SfxResult`
+   * for back-compat.
+   */
   async wait(taskId, opts = {}) {
     const pollInterval = opts.pollInterval ?? DEFAULT_POLL_INTERVAL_MS;
     const timeout = opts.timeout ?? DEFAULT_WAIT_TIMEOUT_MS;
@@ -283,6 +295,32 @@ var TextToMusic = class {
   generate(params) {
     return collectTrack(this.stream(params));
   }
+  /**
+   * Submit an async text-to-music task; poll with
+   * `client.tasks.wait<MusicTaskResult>(task.task_id)`. Required for
+   * `outputFormat: "wav"`. `stream()`/`generate()` remain the streaming path.
+   */
+  async submit(params) {
+    const mode = params.mode ?? "async";
+    if (mode !== "async") {
+      throw new SoniloError('submit() requires mode: "async"');
+    }
+    const form = new FormData();
+    form.set("prompt", params.prompt);
+    form.set("duration", String(params.duration));
+    if (params.segments !== void 0) {
+      form.set("segments", JSON.stringify(params.segments));
+    }
+    form.set("mode", mode);
+    if (params.outputFormat !== void 0) {
+      form.set("output_format", params.outputFormat);
+    }
+    const res = await this.client.request("/v1/text-to-music", {
+      method: "POST",
+      body: form
+    });
+    return await res.json();
+  }
 };
 
 // src/upload.ts
@@ -354,6 +392,54 @@ var VideoToMusic = class {
   generate(params) {
     return collectTrack(this.stream(params));
   }
+  /**
+   * Submit an async video-to-music task; poll its result with
+   * `client.tasks.wait<MusicTaskResult>(task.task_id)`. Required for
+   * `isolateVocals` — the backend rejects vocal isolation on the plain
+   * stream, and it only ever runs in async mode.
+   */
+  async submit(params) {
+    if (params.video === void 0 === (params.videoUrl === void 0)) {
+      throw new SoniloError("Provide exactly one of video or videoUrl");
+    }
+    let mode = params.mode;
+    const needsAsync = params.isolateVocals || params.preserveSpeech || params.ducking !== void 0 || params.outputFormat === "wav";
+    if (mode === void 0) mode = "async";
+    if (needsAsync && mode !== "async") {
+      throw new SoniloError(
+        'isolateVocals/preserveSpeech/ducking/outputFormat "wav" require mode: "async"'
+      );
+    }
+    const form = new FormData();
+    if (params.video !== void 0) {
+      const { blob, filename } = await toUploadBlob(params.video);
+      form.set("video", blob, filename);
+    } else {
+      form.set("video_url", params.videoUrl);
+    }
+    if (params.prompt !== void 0) form.set("prompt", params.prompt);
+    if (params.segments !== void 0) {
+      form.set("segments", JSON.stringify(params.segments));
+    }
+    form.set("mode", mode);
+    if (params.preserveSpeech !== void 0) {
+      form.set("preserve_speech", String(params.preserveSpeech));
+    }
+    if (params.isolateVocals !== void 0) {
+      form.set("isolate_vocals", String(params.isolateVocals));
+    }
+    if (params.outputFormat !== void 0) {
+      form.set("output_format", params.outputFormat);
+    }
+    if (params.ducking !== void 0) {
+      form.set("ducking", String(params.ducking));
+    }
+    const res = await this.client.request("/v1/video-to-music", {
+      method: "POST",
+      body: form
+    });
+    return await res.json();
+  }
 };
 
 // src/resources/textToSfx.ts
@@ -411,8 +497,75 @@ var VideoToSfx = class {
   }
 };
 
+// src/resources/videoToVideoMusic.ts
+var VideoToVideoMusic = class {
+  constructor(client) {
+    this.client = client;
+  }
+  async submit(params) {
+    if (params.video === void 0 === (params.videoUrl === void 0)) {
+      throw new SoniloError("Provide exactly one of video or videoUrl");
+    }
+    const form = new FormData();
+    if (params.video !== void 0) {
+      const { blob, filename } = await toUploadBlob(params.video);
+      form.set("video", blob, filename);
+    } else {
+      form.set("video_url", params.videoUrl);
+    }
+    if (params.prompt !== void 0) form.set("prompt", params.prompt);
+    if (params.preserveSpeech !== void 0) {
+      form.set("preserve_speech", String(params.preserveSpeech));
+    }
+    if (params.isolateVocals !== void 0) {
+      form.set("isolate_vocals", String(params.isolateVocals));
+    }
+    const res = await this.client.request("/v1/video-to-video-music", {
+      method: "POST",
+      body: form
+    });
+    return await res.json();
+  }
+  async generate(params, opts) {
+    const task = await this.submit(params);
+    return this.client.tasks.wait(task.task_id, opts);
+  }
+};
+
+// src/resources/videoToVideoSfx.ts
+var VideoToVideoSfx = class {
+  constructor(client) {
+    this.client = client;
+  }
+  async submit(params) {
+    if (params.video === void 0 === (params.videoUrl === void 0)) {
+      throw new SoniloError("Provide exactly one of video or videoUrl");
+    }
+    const form = new FormData();
+    if (params.video !== void 0) {
+      const { blob, filename } = await toUploadBlob(params.video);
+      form.set("video", blob, filename);
+    } else {
+      form.set("video_url", params.videoUrl);
+    }
+    if (params.prompt !== void 0) form.set("prompt", params.prompt);
+    if (params.segments !== void 0) {
+      form.set("segments", JSON.stringify(params.segments));
+    }
+    const res = await this.client.request("/v1/video-to-video-sfx", {
+      method: "POST",
+      body: form
+    });
+    return await res.json();
+  }
+  async generate(params, opts) {
+    const task = await this.submit(params);
+    return this.client.tasks.wait(task.task_id, opts);
+  }
+};
+
 // src/version.ts
-var VERSION = "0.2.0";
+var VERSION = "0.4.0";
 
 // src/client.ts
 var DEFAULT_BASE_URL = "https://api.sonilo.com";
@@ -436,6 +589,8 @@ var SoniloClient = class {
     this.videoToMusic = new VideoToMusic(this);
     this.textToSfx = new TextToSfx(this);
     this.videoToSfx = new VideoToSfx(this);
+    this.videoToVideoMusic = new VideoToVideoMusic(this);
+    this.videoToVideoSfx = new VideoToVideoSfx(this);
   }
   /**
    * Perform an authenticated request; throws a typed error on non-2xx.