← All models

Fish Audio Drama 3

Audio Available Full comparison ↗

fish-audio/drama-3 · by Fish Audio · neural-tts

Pricing — 1 offering(s)

Fish Audio

About this offering: Fish Audio Drama 3 (model string drama-3), launched October 13, 2026: a text-to-speech model you direct in plain language. Directions go in square brackets before a line, in any language; you can add sounds and pauses where they happen, limit a delivery to specific words with paired tags such as whisper or emphasis, switch language for a phrase (experimental) and pin pronunciations with phoneme tags. It supports multi-speaker dialogue in one request and is available in the API on the same text-to-speech endpoint (select it with the model header) and in Fish Audio's web app. $50 per 1M UTF-8 bytes, with a launch promotion of $10 through November 30, 2026 (Fish Audio pricing page, checked 2026-11-10).

Text characters

  • $10.00 / 1M characters Current 2026-10-26 → 2026-11-30
    Launch promotion: $10 per 1M UTF-8 bytes, 80% off the $50 list price, through November 30, 2026 (until 00:00…

    Launch promotion: $10 per 1M UTF-8 bytes, 80% off the $50 list price, through November 30, 2026 (until 00:00 Pacific on December 1, 08:00 UTC). The promotion's start date is the earliest day Fish Audio's pricing page is known to have shown it, not a confirmed start date. Fish Audio bills per UTF-8 byte: for English and other Latin-script text one character is one byte; for Chinese, Japanese, Korean and other multi-byte scripts the cost per character can be up to 3–4× this rate.

  • $50.00 / 1M characters Future 2026-12-01
    List price, $50 per 1M UTF-8 bytes, from December 1, 2026 when the launch promotion ends. Fish Audio bills…

    List price, $50 per 1M UTF-8 bytes, from December 1, 2026 when the launch promotion ends. Fish Audio bills per UTF-8 byte: for English and other Latin-script text one character is one byte; for Chinese, Japanese, Korean and other multi-byte scripts the cost per character can be up to 3–4× this rate.

Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.

See how Fish Audio Drama 3 fits into a cost-aware routing setup

See how →

Capability profile

How Fish Audio Drama 3 rates across core capability dimensions, with the task-level evidence behind each rating.

voice naturalness Unknown

Fish Audio describes it as built for voice direction and content creation, with lines performed from plain-language directions such as "barely holding it together, forcing a smile". No independent quality rating is recorded yet.

language support Unknown

Fish Audio doesn't publish a language count for Drama 3. Directions can be written in a different language from the script, and switching language mid-line is experimental (Fish Audio's examples cover French, Spanish, German, Japanese, Chinese, Korean and Italian). Phoneme tags cover English, Chinese and Japanese.

voice variety Strong

Multi-speaker dialogue in one request (one voice per speaker), with Fish Audio's public voice library and cloned voices.

cloning support Strong

Works with cloned voices from Fish Audio's voice cloning; for mid-line language switches Fish Audio recommends a voice cloned from a speaker fluent in both languages.

Ratings are estimated — limited independent data is available for this model.

Operator guidance

Choose Drama 3 when the delivery itself has to be directed line by line (emotion, pacing, sounds, whispered words). For general production text-to-speech, Fish Audio's S2.1 Pro covers 83 languages with latency guarantees at $15 per 1M UTF-8 bytes, against Drama 3's $50 list price. The $10 launch promotion ends November 30, 2026, so budget at the list price for anything running past then.

Use cases

Limitations

Citations