Vibeleaderboard
← All Intel
Intel / post

Fish Audio previews Drama 3, a directable speech model

Source
x.com
Date
Fish Audio@FishAudio

Drama 3 doesn't just read lines. It brings the drama. Direct the performance in plain language: tone, emotion, pacing, character. It can even shift emotion mid-line, no retake needed. Try the preview now in the API with 'drama-3-preview'.

Context

Fish Audio said in an X post from its account that its Drama 3 speech , which turns written lines into spoken audio, lets a developer direct a performance in plain language, covering tone, emotion, pacing and character. It claims the model can shift emotion in the middle of a line without a retake, and says a preview is available in the company's under the name drama-3-preview.

The attached demo video could be checked only through an automatic transcript of its audio. That transcript is a short comic monologue in which a speaker voices a device as its battery falls from 20% to 1%, moving from forced calm ("That's fine. Totally fine.") to pleading and then relief ("Oh, thank God") at the end. A transcript records the words, not how they are delivered or what appears on screen, so it cannot show how the performance was directed or how well the emotional shifts land, and the post itself offers no listening tests, comparisons or limits. An earlier Fish Audio post introduced the Drama 3 preview and offered API keys to people who commented on it.

Terms in this piece · Glossary
  • Model — A computational system that uses learned patterns to produce an output from information supplied to it.
  • API — Application programming interface: the agreed set of operations, inputs and outputs through which one piece of software uses another.
More from Fish Audio
Recommended reads
Comments

Checking sign-in…

Loading comments…