Fish Audio Drama 3 - Same Scripts vs Eleven v4
Fish Audio Drama 3, tested: Fish’s own demo scripts run through Eleven v4, with real audio, plus price ($10 per million until Nov 12), license and syntax.
Last updated: 2026-10-19
Fish Audio Drama 3 is Fish Audio’s new directable text-to-speech model, released on October 13, 2026: you write stage directions in square brackets, anywhere in a line, in 83 languages. We took the six demo scripts Fish published with the launch and ran each one through ElevenLabs’ Eleven v4, word for word, so you can compare the same scripts on both models. Fish’s own Fish Audio Drama 3 takes play on Fish’s launch post; the Eleven v4 audio below is ours, one take each, unedited.
Fish Audio Drama 3 vs Eleven v4 at a glance
| Fish Audio Drama 3 | Eleven v4 | |
|---|---|---|
| Released | October 13, 2026 | September 28, 2026 |
| Languages | 83 | 90+ |
| How you direct it | Free-form directions in [brackets], at the start or mid-line, in any language | Audio tags in [brackets] such as [whispers], [crying], [laughing] |
| Inline markup | Paired tags such as <emphasis>…</emphasis> and <whisper>…</whisper> | None; tags apply from where they appear |
| API price | $50 per million characters; $10 per million until November 12, 2026 | $0.08 per 1,000 characters ($80 per million); v4 Turbo $0.04 |
| Weights | Open, on Hugging Face (fishaudio/drama-3) | Closed |
| Commercial use | Free under US$500,000 annual revenue; a license above that or to resell it as a service | On ElevenLabs’ paid plans |
Prices are the vendors’ own list prices, checked on October 19, 2026. Fish also reports that Fish Audio Drama 3 scores 93.2% on Artificial Analysis’ English pronunciation-robustness test, against 88.1% for its previous model, S2.1 Pro.
How we tested Fish Audio Drama 3’s scripts
We do not have a Fish Audio key, so we did not generate Fish Audio Drama 3 audio ourselves. Fish published its takes of these exact scripts on its launch post, and that is the Drama 3 side of this comparison. For the Eleven v4 side we rendered each script twice, with the same voice both times:
- Verbatim: Fish’s script exactly as published, free-form directions and all, to see whether Eleven v4 copes with Drama 3-style writing.
- With v4 tags: the same lines with Eleven v4’s own audio tags, the way you would write them for ElevenLabs.
Every take was then transcribed with ElevenLabs Scribe to check whether any direction was spoken aloud. One take per version, no retries, no editing.
The Fish Audio Drama 3 demo scripts, side by side
1. A forced smile
[barely holding it together, forcing a smile] I'm fine. [shaky breath] Really. [small laugh] [whispering] Go on without me.
Drama 3: Fish’s own take, on the launch post
Eleven v4, verbatim script:
[shaky breath] Really. [small laugh] [whispering] Go on without me.Eleven v4, with v4 tags ([dejected] I'm fine. [sighs] Really. [chuckles] [whispers] Go on without me.):
[dejected] I'm fine. [sighs] Really. [chuckles] [whispers] Go on without me.2. Choking up
[choked up, crying quietly] I left his jacket by the door. [voice breaking] I keep thinking he'll walk in and ask for it.
Drama 3: Fish’s own take, on the launch post
Eleven v4, verbatim script:
[voice breaking] I keep thinking he'll walk in and ask for it.Eleven v4, with v4 tags ([crying] … [sobbing] …):
[crying] I left his jacket by the door. [sobbing] I keep thinking he'll walk in and ask for it.3. Holding back anger
[through gritted teeth, keeping his voice down] I asked you to do one thing tonight. [slow, dangerously quiet] Just one.
Drama 3: Fish’s own take, on the launch post
Eleven v4, verbatim script. In this take Eleven v4 said “Just one” twice:
Eleven v4, with v4 tags ([angry] … [menacing] Just one.):
[angry] I asked you to do one thing tonight. [menacing] Just one.4. Laughing despite herself
[straight-faced, trying very hard to stay serious] No, I love it, it really suits you. [cracking up despite herself] I'm sorry, it's just a very brave haircut.
Drama 3: Fish’s own take, on the launch post
Eleven v4, verbatim script. The free-form “cracking up” direction produced a real laugh:
[cracking up despite herself] I'm sorry, it's just a very brave haircut.Eleven v4, with v4 tags ([earnestly] … [laughing] …):
[earnestly] No, I love it, it really suits you. [laughing] I'm sorry, it's just a very brave haircut.5. Don’t open that door
[breathless, speaking fast] Just turn off the lights. Stay away from the windows. [voice drops to a terrified whisper] And whatever you hear out there… [slow, deadly serious] <emphasis>do not open that door.</emphasis>
Drama 3: Fish’s own take, on the launch post
Eleven v4, verbatim script, <emphasis> markup included:
[voice drops to a terrified whisper] And whatever you hear out there… [slow, deadly serious] <emphasis>do not open that door.</emphasis>Eleven v4, with v4 tags ([fearful] … [whispers] … [solemnly] …):
[fearful] Just turn off the lights. Stay away from the windows. [whispers] And whatever you hear out there… [solemnly] do not open that door.6. A direction written in Chinese
Drama 3 accepts directions in any language. Fish’s example writes the direction in Chinese and the line in English:
[像是在强忍着眼泪，却努力笑着] I'm fine. Really. Go on without me.
Drama 3: Fish’s own take, on the launch post
Eleven v4, verbatim script:
What the Fish Audio Drama 3 scripts showed about Eleven v4
- No direction was ever spoken. Across all 11 Eleven v4 takes, the Scribe transcripts contain only the dialogue: none of Fish’s free-form directions, not the Chinese one, and not the
<emphasis>markup. - Some free-form directions were acted. “Cracking up despite herself” produced an audible laugh, and the forced-smile script got a small chuckle in both versions. Writing in Drama 3’s style does not break Eleven v4.
- One slip. The verbatim take of script 3 repeated “Just one”. The tagged take did not. With one take per version we cannot say how often that happens.
- What we cannot measure for you: which performance is more convincing. That is a listening judgment, so play Fish’s takes and ours one after the other.
Directing emotion: Fish Audio Drama 3 vs Eleven v4
Fish Audio Drama 3 treats anything in square brackets as a stage direction you write in your own words, “[voice drops to a terrified whisper]” as readily as “[whispering]”, and lets paired tags such as <emphasis> cover a few words. Eleven v4 documents a set of audio tags, [whispers], [laughs], [sighs], [crying] and the like, which apply from the point where they appear. If you move a script from one to the other, the safest edit is to keep each direction short and put it right before the line it colours.
Price and license
At list price Fish Audio Drama 3 costs $50 per million characters, and $10 per million until November 12, 2026. Eleven v4 costs $80 per million characters on the ElevenLabs API, and v4 Turbo $40. The bigger difference is the license: Drama 3’s weights are open and free for commercial use only while you and your affiliates earn under US$500,000 a year, and offering it as a TTS or voice service of your own needs a separate license from Fish.
Which to use
- Choose Fish Audio Drama 3 if you are under the revenue threshold, want the lowest price per character this month, want to self-host the weights, or like writing directions in plain words.
- Choose Eleven v4 if you need its 90+ languages, a large ready-made voice library, or a commercial footing that does not depend on your revenue.
VoiceSmiths runs its audio drama studio on Eleven v4: paste a story, the characters are found and cast, and every line gets a direction automatically. Finding the cast is free. If Eleven v4 is new to you, start with the Eleven v4 vs v3 test.
FAQ
What is Fish Audio Drama 3? Fish Audio Drama 3 is a text-to-speech model from Fish Audio, released on October 13, 2026, built for directed, emotional speech. You write stage directions in square brackets, at the start of a line or in the middle of it, and it covers 83 languages. It runs on fish.audio, in Fish’s apps and through the API under the model name drama-3.
How much does Fish Audio Drama 3 cost? Fish lists the Drama 3 API at $50 per million characters, cut to $10 per million from October 13 to November 12, 2026. For comparison, ElevenLabs lists Eleven v4 at $0.08 per 1,000 characters ($80 per million) and v4 Turbo at $0.04 (checked October 19, 2026).
Is Fish Audio Drama 3 free for commercial use? According to Fish’s launch post, Drama 3 is under the Fish Audio Community License v1.0: free for research and non-commercial use, and free for commercial use if you and your affiliates earn under US$500,000 a year. Above that, or to offer Drama 3 as a TTS, voice-cloning or inference service, you need a license from Fish. The Hugging Face model page also shows a research-license tag, so read the LICENSE file before relying on it.
Does Eleven v4 understand Fish Audio Drama 3-style directions? In our test it never read them aloud: none of Fish’s free-form directions, a direction written in Chinese, or the <emphasis> markup turned up in the speech-to-text transcripts. Some were acted, too: “cracking up despite herself” produced a real laugh. Eleven v4’s own audio tags such as [whispers], [crying] and [laughing] are the documented way to direct it.
Is Fish Audio Drama 3 better than Eleven v4? It depends on what you need, and one take per script cannot rank two models. Drama 3 is cheaper per character, has open weights and a very flexible direction syntax. Eleven v4 covers 90+ languages, has a large voice library and no revenue threshold on paid plans. Listen to Fish’s own takes and the Eleven v4 renders on this page side by side and judge the delivery yourself.
Fish Audio and Drama 3 are trademarks of their owners. VoiceSmiths is independent and not affiliated with Fish Audio or ElevenLabs.