Fish Audio Drama 3 Is Here: The Most Controllable TTS Model Ever

Summary
This video is an official preview teaser by Fish Audio showcasing their text-to-speech model, Drama 3. It demonstrates directable voice acting and fine-grained emotional and delivery control across multiple languages using natural language brackets and inline markup tags.

What is shown

• [00:02] Title card introducing “Fish Audio Drama 3 Preview: A voice you can direct.”
• [00:04] – [00:09] Feature highlights presented on screen: “Add a little [ performance ]”, “Describe it in your own words”, and “Choose exactly which words change.”
• [00:10] – [00:20] English voice demo (“Mila”): Audio generation showing rapid mid-sentence shifts guided by prompt tags ([breathless, speaking fast, urgent], [voice drops to a terrified whisper], and [slow, deadly serious] <emphasis> do not open that door. </emphasis>).
• [00:21] – [00:27] Japanese voice demo (“陽翔 Haruto”): Audio delivery demonstrating emotional nuance using tags ([surprised, breath catching] and [softening into a smile] [pause]).
• [00:28] – [00:34] French voice demo (“Thérèse”): Conversational speech shifting into whispered delivery using inline tags ([leaning in, conspiratorial] and <whisper> entre nous </whisper>).
• [00:35] – [00:40] Outro card announcing availability via Fish Audio Creative Studio and API with early access rolling out.

Claims & numbers

• The video claims users can direct voice performance in their own words and specify exact word-level changes.
• Fish Audio claims early access is rolling out for Drama 3 across Fish Audio Creative Studio and their API.

Notable quotes

• “Just turn off the lights. Stay away from the windows. And whatever you hear out there... do not open that door.” — Mila voice demo [00:11]
• “え、本当に…？ 戻ってきてくれるなんて、思ってなかった。” — Haruto voice demo [00:21]
• “J’ai un truc à te dire... Ça reste entre nous, d’accord ?” — Thérèse voice demo [00:29]

Assessment
This is an official product teaser and feature showcase illustrating Drama 3’s tagged prompt interface and speech generation. While the audio samples cleanly reflect the inline text directions, the video presents pre-rendered showcase clips rather than an unedited live interactive interface demonstration.

Described by gemini-3.8-flash on 2026-10-13 from the video’s audio and frames.