S2.1 Pro
A previously released speech generation model from Fish Audio, the predecessor to Drama 3, used as a comparison baseline in benchmarks such as pronunciation robustness.
Timeline (last 90 days)
Drama 3 scored 93.2% on Artificial Analysis' pronunciation robustness benchmark, ranking first, up from 88.1% for the previous generation S2.1 Pro.
On exact sequences such as file paths and email addresses, Drama 3's accuracy improved from 79.4% to 86.3%.
In Drama 3 API calls, if the model header is set incorrectly, the request does not fail but silently falls back to the S2.1 Pro model.
S2.1 Pro employs cross-lingual voice cloning technology that decouples a speaker's timbral characteristics from the phonemic system of the language, allowing the same voice ID to maintain a consistent vocal identity across different languages.
The core reason for choosing S2.1 Pro was its ability to maintain the same voice when switching between different languages such as English and Hindi.
The voice-approval errors were not the fault of the voice model — the audio performance was perfect; the error lay in the fact that the content of the statement itself was wrong.
The entire voice integration requires only a single function: pass in text and a voice ID, return audio, with no retraining or second model needed.
Fish Audio's S2.1 Pro developer API is free to access through the end of November, subject to a fair use policy.
All Facts (8)
Drama 3 scored 93.2% on Artificial Analysis' pronunciation robustness benchmark, ranking first, up from 88.1% for the previous generation S2.1 Pro.
50%UnverifiedOn exact sequences such as file paths and email addresses, Drama 3's accuracy improved from 79.4% to 86.3%.
50%UnverifiedIn Drama 3 API calls, if the model header is set incorrectly, the request does not fail but silently falls back to the S2.1 Pro model.
50%UnverifiedS2.1 Pro employs cross-lingual voice cloning technology that decouples a speaker's timbral characteristics from the phonemic system of the language, allowing the same voice ID to maintain a consistent vocal identity across different languages.
50%UnverifiedThe core reason for choosing S2.1 Pro was its ability to maintain the same voice when switching between different languages such as English and Hindi.
50%UnverifiedThe voice-approval errors were not the fault of the voice model — the audio performance was perfect; the error lay in the fact that the content of the statement itself was wrong.
50%UnverifiedThe entire voice integration requires only a single function: pass in text and a voice ID, return audio, with no retraining or second model needed.
50%UnverifiedFish Audio's S2.1 Pro developer API is free to access through the end of November, subject to a fair use policy.
50%