@nodaro/sdk 1.1.1 → 1.2.0

package/dist/index.d.cts

@@ -3402,18 +3402,22 @@ declare class VoicesResource {
      * ElevenLabs UUID) OR a {@link VoiceChangerProVoice} object carrying per-voice
      * ElevenLabs speech-to-speech settings (stability / similarityBoost / style /
      * useSpeakerBoost / `seed`) plus a loudness `volumeMode` (and a manual
-     * `volume`). A per-voice `seed` makes that speaker's recast reproducible.
+     * `volume`) — OR `null`, meaning keep this speaker's original voice (requires
+     * a platform running cloud-plugins with keep-slot support). At least one
+     * entry must be non-null. A per-voice `seed` makes that speaker's recast
+     * reproducible.
      *
      * Pass `audioUrl` for audio-only recast or `videoUrl` to recast the audio
      * track of a video clip (the server demuxes, recasts, and remuxes).
      *
-     * Voice and music are ALWAYS separated first — ElevenLabs only ever sees the
-     * isolated vocal stem, never the music bed. `preserveBackground` (default
-     * `true`) only controls whether that music/instrumental stem is mixed back
-     * under the new voices; set it `false` for a clean voice-only result.
-     * `separationQuality` selects the demucs model used for that split: `"fast"`
-     * (default, htdemucs — preserves more of the voice) or `"best"` (htdemucs_ft —
-     * finer separation). `removeBackgroundNoise` additionally denoises the result.
+     * Voice and music are ALWAYS separated first — before recasting, the source
+     * is split into an isolated vocal stem and a music/SFX stem.
+     * `preserveBackground` (default `true`) only controls whether that
+     * music/instrumental stem is mixed back under the new voices; set it `false`
+     * for a clean voice-only result. `separationQuality` selects the quality of
+     * the voice/music separation: `"fast"` (default, quicker — preserves more of
+     * the voice) or `"best"` (finer voice/music separation).
+     * `removeBackgroundNoise` additionally denoises the result.
      * `musicVolumeMode` sets the level of that preserved background (only relevant
      * when `preserveBackground` is on): `"match"` (default) keeps the original
      * level, `"normalize"` loudnorms it, `"manual"` uses `musicVolume`%.
@@ -3452,8 +3456,8 @@ type VoiceChangerProVoice = string | {
     seed?: number;
     /**
      * Loudness handling for this recast voice. `"match"` (default) matches the
-     * original speaker's loudness; `"normalize"` applies EBU R128 loudnorm;
-     * `"manual"` uses `volume` as a percentage.
+     * original speaker's loudness; `"normalize"` applies loudness
+     * normalization; `"manual"` uses `volume` as a percentage.
      */
     volumeMode?: "match" | "normalize" | "manual";
     /** Manual output volume as a percentage (0–200). Consulted only when `volumeMode === "manual"`. */
@@ -3468,9 +3472,12 @@ interface VoiceChangerProInput {
     /**
      * Voices in speaker-detection order. Speaker N is mapped to `orderedVoices[N]`;
      * speakers beyond the array keep their original voice. Each entry is a bare
-     * voice id OR a {@link VoiceChangerProVoice} object with per-voice settings.
+     * voice id OR a {@link VoiceChangerProVoice} object with per-voice settings —
+     * OR `null`, meaning keep this speaker's original voice (requires a platform
+     * running cloud-plugins with keep-slot support). At least one entry must be
+     * non-null.
      */
-    orderedVoices: Array<VoiceChangerProVoice>;
+    orderedVoices: Array<VoiceChangerProVoice | null>;
     /** Model to use for speech-to-speech. Defaults to the server-configured default when omitted. */
     model?: string;
     /**
@@ -3480,8 +3487,8 @@ interface VoiceChangerProInput {
      */
     preserveBackground?: boolean;
     /**
-     * Demucs model used to split voice from music. `"fast"` (default, htdemucs —
-     * preserves more of the voice) or `"best"` (htdemucs_ft — finer separation).
+     * Quality of the voice/music separation. `"fast"` (default, quicker —
+     * preserves more of the voice) or `"best"` (finer voice/music separation).
      */
     separationQuality?: "fast" | "best";
     /** Strip background noise for a clean voice-only result. */
@@ -3490,8 +3497,8 @@ interface VoiceChangerProInput {
      * Level of the preserved background music / SFX stem in the final mix. Only
      * relevant when `preserveBackground` is on (otherwise there is no background to
      * level). `"match"` (default) leaves the separated instrumental at its original
-     * level; `"normalize"` applies EBU R128 loudnorm; `"manual"` sets its level to
-     * `musicVolume`%.
+     * level; `"normalize"` applies loudness normalization; `"manual"` sets its
+     * level to `musicVolume`%.
      */
     musicVolumeMode?: "match" | "normalize" | "manual";
     /** Background music level as a percentage (0–200). Consulted only when `musicVolumeMode === "manual"`. */

package/dist/index.d.ts

@@ -3402,18 +3402,22 @@ declare class VoicesResource {
      * ElevenLabs UUID) OR a {@link VoiceChangerProVoice} object carrying per-voice
      * ElevenLabs speech-to-speech settings (stability / similarityBoost / style /
      * useSpeakerBoost / `seed`) plus a loudness `volumeMode` (and a manual
-     * `volume`). A per-voice `seed` makes that speaker's recast reproducible.
+     * `volume`) — OR `null`, meaning keep this speaker's original voice (requires
+     * a platform running cloud-plugins with keep-slot support). At least one
+     * entry must be non-null. A per-voice `seed` makes that speaker's recast
+     * reproducible.
      *
      * Pass `audioUrl` for audio-only recast or `videoUrl` to recast the audio
      * track of a video clip (the server demuxes, recasts, and remuxes).
      *
-     * Voice and music are ALWAYS separated first — ElevenLabs only ever sees the
-     * isolated vocal stem, never the music bed. `preserveBackground` (default
-     * `true`) only controls whether that music/instrumental stem is mixed back
-     * under the new voices; set it `false` for a clean voice-only result.
-     * `separationQuality` selects the demucs model used for that split: `"fast"`
-     * (default, htdemucs — preserves more of the voice) or `"best"` (htdemucs_ft —
-     * finer separation). `removeBackgroundNoise` additionally denoises the result.
+     * Voice and music are ALWAYS separated first — before recasting, the source
+     * is split into an isolated vocal stem and a music/SFX stem.
+     * `preserveBackground` (default `true`) only controls whether that
+     * music/instrumental stem is mixed back under the new voices; set it `false`
+     * for a clean voice-only result. `separationQuality` selects the quality of
+     * the voice/music separation: `"fast"` (default, quicker — preserves more of
+     * the voice) or `"best"` (finer voice/music separation).
+     * `removeBackgroundNoise` additionally denoises the result.
      * `musicVolumeMode` sets the level of that preserved background (only relevant
      * when `preserveBackground` is on): `"match"` (default) keeps the original
      * level, `"normalize"` loudnorms it, `"manual"` uses `musicVolume`%.
@@ -3452,8 +3456,8 @@ type VoiceChangerProVoice = string | {
     seed?: number;
     /**
      * Loudness handling for this recast voice. `"match"` (default) matches the
-     * original speaker's loudness; `"normalize"` applies EBU R128 loudnorm;
-     * `"manual"` uses `volume` as a percentage.
+     * original speaker's loudness; `"normalize"` applies loudness
+     * normalization; `"manual"` uses `volume` as a percentage.
      */
     volumeMode?: "match" | "normalize" | "manual";
     /** Manual output volume as a percentage (0–200). Consulted only when `volumeMode === "manual"`. */
@@ -3468,9 +3472,12 @@ interface VoiceChangerProInput {
     /**
      * Voices in speaker-detection order. Speaker N is mapped to `orderedVoices[N]`;
      * speakers beyond the array keep their original voice. Each entry is a bare
-     * voice id OR a {@link VoiceChangerProVoice} object with per-voice settings.
+     * voice id OR a {@link VoiceChangerProVoice} object with per-voice settings —
+     * OR `null`, meaning keep this speaker's original voice (requires a platform
+     * running cloud-plugins with keep-slot support). At least one entry must be
+     * non-null.
      */
-    orderedVoices: Array<VoiceChangerProVoice>;
+    orderedVoices: Array<VoiceChangerProVoice | null>;
     /** Model to use for speech-to-speech. Defaults to the server-configured default when omitted. */
     model?: string;
     /**
@@ -3480,8 +3487,8 @@ interface VoiceChangerProInput {
      */
     preserveBackground?: boolean;
     /**
-     * Demucs model used to split voice from music. `"fast"` (default, htdemucs —
-     * preserves more of the voice) or `"best"` (htdemucs_ft — finer separation).
+     * Quality of the voice/music separation. `"fast"` (default, quicker —
+     * preserves more of the voice) or `"best"` (finer voice/music separation).
      */
     separationQuality?: "fast" | "best";
     /** Strip background noise for a clean voice-only result. */
@@ -3490,8 +3497,8 @@ interface VoiceChangerProInput {
      * Level of the preserved background music / SFX stem in the final mix. Only
      * relevant when `preserveBackground` is on (otherwise there is no background to
      * level). `"match"` (default) leaves the separated instrumental at its original
-     * level; `"normalize"` applies EBU R128 loudnorm; `"manual"` sets its level to
-     * `musicVolume`%.
+     * level; `"normalize"` applies loudness normalization; `"manual"` sets its
+     * level to `musicVolume`%.
      */
     musicVolumeMode?: "match" | "normalize" | "manual";
     /** Background music level as a percentage (0–200). Consulted only when `musicVolumeMode === "manual"`. */

package/dist/index.js

@@ -1522,18 +1522,22 @@ var VoicesResource = class {
    * ElevenLabs UUID) OR a {@link VoiceChangerProVoice} object carrying per-voice
    * ElevenLabs speech-to-speech settings (stability / similarityBoost / style /
    * useSpeakerBoost / `seed`) plus a loudness `volumeMode` (and a manual
-   * `volume`). A per-voice `seed` makes that speaker's recast reproducible.
+   * `volume`) — OR `null`, meaning keep this speaker's original voice (requires
+   * a platform running cloud-plugins with keep-slot support). At least one
+   * entry must be non-null. A per-voice `seed` makes that speaker's recast
+   * reproducible.
    *
    * Pass `audioUrl` for audio-only recast or `videoUrl` to recast the audio
    * track of a video clip (the server demuxes, recasts, and remuxes).
    *
-   * Voice and music are ALWAYS separated first — ElevenLabs only ever sees the
-   * isolated vocal stem, never the music bed. `preserveBackground` (default
-   * `true`) only controls whether that music/instrumental stem is mixed back
-   * under the new voices; set it `false` for a clean voice-only result.
-   * `separationQuality` selects the demucs model used for that split: `"fast"`
-   * (default, htdemucs — preserves more of the voice) or `"best"` (htdemucs_ft —
-   * finer separation). `removeBackgroundNoise` additionally denoises the result.
+   * Voice and music are ALWAYS separated first — before recasting, the source
+   * is split into an isolated vocal stem and a music/SFX stem.
+   * `preserveBackground` (default `true`) only controls whether that
+   * music/instrumental stem is mixed back under the new voices; set it `false`
+   * for a clean voice-only result. `separationQuality` selects the quality of
+   * the voice/music separation: `"fast"` (default, quicker — preserves more of
+   * the voice) or `"best"` (finer voice/music separation).
+   * `removeBackgroundNoise` additionally denoises the result.
    * `musicVolumeMode` sets the level of that preserved background (only relevant
    * when `preserveBackground` is on): `"match"` (default) keeps the original
    * level, `"normalize"` loudnorms it, `"manual"` uses `musicVolume`%.