@omg-dev/media 0.4.29 → 0.4.31

package/dist/index.mjs

@@ -115,16 +115,16 @@ async function generateSpeech(opts) {
 /**
 * Speech → text. Submits, polls to completion, and returns the transcript.
 *
-*   const { text } = await transcribe({ audioUrl, durationSeconds: 42 })
+*   const { text } = await transcribe({ audioUrl })
 */
 async function transcribe(opts) {
 	const { audioUrl, durationSeconds, model, input, ...rest } = opts;
 	const job = await generate({
-		model: model ?? "elevenlabs/scribe_v1",
+		model: model ?? "elevenlabs/scribe_v2",
 		input: {
+			...input,
 			audio_url: audioUrl,
-			duration_seconds: durationSeconds,
-			...input
+			duration_seconds: durationSeconds
 		},
 		...rest
 	});

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@omg-dev/media",
-  "version": "0.4.29",
+  "version": "0.4.31",
   "type": "module",
   "exports": {
     ".": {

package/src/index.ts

@@ -205,13 +205,9 @@ export async function generateSpeech(
 export interface TranscribeOptions extends Omit<GenerateOptions, "model" | "input"> {
   /** Public URL of the audio to transcribe (e.g. an uploaded file's URL). */
   audioUrl: string
-  /**
-   * Length of the audio in seconds — REQUIRED, it is what you're billed on
-   * ($0.22/audio-hour pass-through, 60s minimum). Understating it undercharges;
-   * read it off the file/recorder before submitting.
-   */
-  durationSeconds: number
-  /** Curated model id (default "elevenlabs/scribe_v1"). */
+  /** Optional UI hint; the host measures the real duration before billing. */
+  durationSeconds?: number
+  /** Curated model id (default "elevenlabs/scribe_v2"). */
   model?: string
   /** Extra model params merged into the input (language_code…). */
   input?: Record<string, unknown>
@@ -220,15 +216,15 @@ export interface TranscribeOptions extends Omit<GenerateOptions, "model" | "inpu
 /**
  * Speech → text. Submits, polls to completion, and returns the transcript.
  *
- *   const { text } = await transcribe({ audioUrl, durationSeconds: 42 })
+ *   const { text } = await transcribe({ audioUrl })
  */
 export async function transcribe(
   opts: TranscribeOptions,
 ): Promise<{ text: string; job: Job }> {
   const { audioUrl, durationSeconds, model, input, ...rest } = opts
   const job = await generate({
-    model: model ?? "elevenlabs/scribe_v1",
-    input: { audio_url: audioUrl, duration_seconds: durationSeconds, ...input },
+    model: model ?? "elevenlabs/scribe_v2",
+    input: { ...input, audio_url: audioUrl, duration_seconds: durationSeconds },
     ...rest,
   })
   const text = job.results?.[0]?.text