This is not a request for a new feature.

This functionality already existed and worked correctly until recently.

As a paying Agent Opus customer, I have successfully created numerous AI-generated music videos by uploading purely instrumental audio files (no vocals, no narration, no lyrics) and providing only a brief prompt describing the visuals.

That workflow was one of the primary reasons I subscribed to Agent Opus.

Recently, the behavior changed.

Instrumental audio files now fail because Agent Opus appears to require a transcript before processing. Since instrumental music contains no spoken words whatsoever, there is nothing to transcribe. As a result, an entire workflow that previously functioned correctly is now unusable.

Current Behavior

Upload purely instrumental audio.

Enter a visual prompt.

Agent Opus refuses to continue because it requires a transcript.

Error shown:

"Audio upload failed. Upload another file or remove this clip."

Previous Behavior

The exact same instrumental audio files uploaded successfully.

No transcript was required.

Agent Opus generated complete AI videos directly from instrumental music.

Why this matters

This is not a niche workflow.

Many users—including:

composers

film composers

orchestral musicians

producers

electronic artists

ambient artists

soundtrack creators

classical musicians

create music that intentionally contains no speech.

Requiring a transcript for content that has no words makes Agent Opus unusable for an entire category of creators.

Please investigate whether this is:

a regression introduced in a recent update,

an account-specific issue,

or an intentional product change.

If intentional, please explain why previously supported workflows were removed.

If unintentional, please restore the prior behavior.

Engineering should be able to reproduce this easily by uploading any instrumental audio file with no spoken content and attempting to generate a video.

I have already provided support with reproduction steps, screenshots, and conversation history documenting the regression.

Please restore support for instrumental audio generation without requiring transcripts.