AI news · October 15, 2026
Fish Audio releases Drama 3 for expressive speech
Expressive speech is becoming cheaper to test: Drama 3 combines open weights, an API and a launch price that supports small creator experiments.
- pronunciation score claim
- 93.2%
- launch API price
- $10/M
- languages
- 83
Fish Audio announced Drama 3 on October 13 for its website, mobile apps and API. The company reports a 93.2% score on the Artificial Analysis pronunciation-robustness benchmark, support for 83 languages and open weights on Hugging Face. The API model is named drama-3, and the launch price is listed at $10 per million characters, reduced from a stated $50 list price. The announcement also describes limited commercial use under a revenue threshold and separate licensing for larger services.
The release gives creators two useful paths: a hosted API for fast testing and weights for teams that need more control. The benchmark is a vendor claim, so compare it with the failure cases that matter in your content. Check pronunciation, speaker consistency, long-form stability and rights handling before replacing an existing voice workflow.
What you can do with it
Try it: send the same 60-second script through the Drama 3 API and one current voice provider. Score names, numbers, pauses, emotion, pronunciation and total characters. Keep the cheaper route only if listeners and downstream editing time agree with the benchmark result.
Our take
This is a practical creator release because it lowers the cost of a controlled voice comparison. The strongest signal is whether open weights or the API remove enough editing work to change unit economics. Check the licence before commercial launch.
Links Try itModel weights
Source: Fish Audio — Made With Models writes the brief; the reporting is theirs.