On Windows with Premiere Pro Beta 27.0, this direct UXP call appears to ignore the requested language:
const ok = await ppro.Transcript.transcribeClipProjectItem(
clip,
{ languageCode: "de-DE" }
);
const transcript = JSON.parse(await ppro.Transcript.exportToJSON(clip));
With German audio and Premiere’s default transcription language set to English:
ok is true, but transcript.language is "en-us" and the text is incorrect.
- Enabling language auto detection in Preferences did not resolve this.
- Manually transcribing the same audio with German selected produces correct German text and
"de-de".
This reproduces directly in the UXP debugger, without any wrapper.
Could you also clarify the intended semantics of languageCode: does it select the recognition language, provide a hint that automatic language detection may override, or only set the language metadata in the exported transcript? How should it interact with the default language and automatic language detection preferences?
Hey @mamoworld
It looks like we have the option here incorrectly documented. Can you try using language instead of languageCode?
Could you also clarify the intended semantics of languageCode: does it select the recognition language, provide a hint that automatic language detection may override, or only set the language metadata in the exported transcript? How should it interact with the default language and automatic language detection preferences?
It sets which language to use for the transcription. The interaction between these settings is:
- By default use transcription preferences
- The
language option overrides the transcription preference setting
- If the “auto-detect language” preference is enabled, that overrides both the
language option and any default preferences.
Thanks, @clegleiter —that explains it! Switching the native option from languageCode to language fixed our test on Windows Premiere Beta 27.0.
With auto-detection disabled and English set as the default, { language: "de-DE" } produced an accurate German transcript (de-de). Omitting the option used English (en-us) and produced incorrect recognition of the same German speech.
Thanks also for clarifying the precedence of auto-detection, the API option, and the default preference!
3 Likes