{"schema":"postcutoff/model@1","as_of":"2026-10-14T00:22:00+02:00","url":"https://postcutoff.com/m/fish-audio-drama-3/","md":"https://postcutoff.com/m/fish-audio-drama-3/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"fish-audio-drama-3","name":"Fish Audio Drama 3","org":"Fish Audio","family":"Fish Audio Drama","released":"2026-10-13","status":"current","type":"audio/speech","modality_in":["text","audio"],"modality_out":["audio"],"open_weights":true,"model_license":"Fish Audio Community License v1.0 (research and non-commercial free; commercial use only under USD 500,000 annual revenue; no hosted services without a separate agreement)","context_window":null,"max_output":null,"knowledge_cutoff":null,"pricing":{"per_1m_utf8_bytes":2,"unit":"USD per 1M UTF-8 bytes on OpenRouter (2026-10-13). Fish Audio's own pricing page did not list Drama 3 yet; its press release says API usage is 80% off for the first 30 days, so this may be a launch price (list price unverified)","source":"https://openrouter.ai/fish-audio/drama-3"},"price_line":"$2 per 1M UTF-8 bytes","access":[{"provider":"Fish Audio API","model_id":"drama-3","endpoint":"https://api.fish.audio/v1/tts (model passed in `model` header; a missing or misspelled header falls back to s2.1-pro)","docs":"https://docs.fish.audio/features/text-to-speech/drama-3"},{"provider":"Fish Audio API (preview build)","model_id":"drama-3-preview","endpoint":"https://api.fish.audio/v1/tts","docs":"https://docs.fish.audio/developer-guide/getting-started/changelog"},{"provider":"OpenRouter","model_id":"fish-audio/drama-3","endpoint":"https://openrouter.ai/api/v1/audio/speech","url":"https://openrouter.ai/fish-audio/drama-3"},{"provider":"Hugging Face","model_id":"fishaudio/drama-3","url":"https://huggingface.co/fishaudio/drama-3"},{"provider":"GitHub","url":"https://github.com/fishaudio/fish-speech"},{"provider":"SGLang-Omni (self-hosting)","url":"https://sgl-project.github.io/sglang-omni/cookbook/fishaudio_drama3.html"},{"provider":"Web app","url":"https://fish.audio/app/text-to-speech/?version=drama-3"}],"capabilities":[{"name":"Plain-language direction with word-level tags","detail":"Free-text direction in square brackets (any language), point tags such as [sighs], and paired tags <whisper>, <emphasis>, <soft>, <fast>, <slow>, <stress>, <forceful> that apply to specific words.","first":false,"discovered":"launch","source":"https://github.com/fishaudio/docs/blob/main/features/text-to-speech/drama-3.mdx"},{"name":"Editing by infilling","detail":"Replaces a single word in an existing take while keeping the surrounding audio; vendor-reported Ming-Freeform WER 3.49 (English) and 3.69 (Chinese) against 4.58 and 8.64 for Cartesia Sonic-3 infill.","first":false,"discovered":"launch","source":"https://github.com/fishaudio/fish-speech/blob/main/docs/assets/Fish_Audio_Drama_3_Technical_Report.pdf"},{"name":"Multi-turn revision","detail":"A note in the next turn (e.g. 'gentler, and a bit sadder') regenerates the previous take with it in context; vendor-reported style accuracy 76.5 → 81.8 over three rounds on Step-Audio-Edit.","first":false,"discovered":"launch","source":"https://github.com/fishaudio/fish-speech/blob/main/docs/assets/Fish_Audio_Drama_3_Technical_Report.pdf"},{"name":"Multi-character scenes","detail":"Speaker tags <|speaker:0|> to <|speaker:15|>: up to 16 characters in one text, with voice cloning from references; long text is chunked and streamed.","first":false,"discovered":"launch","source":"https://huggingface.co/fishaudio/drama-3"},{"name":"Large open-weights speech model with fast serving","detail":"35B-parameter MoE backbone, 3B active per token; vendor-reported 88.9 ms median time to first audio and RTF 0.131 for one stream on one NVIDIA H200 with SGLang-Omni.","first":false,"discovered":"launch","source":"https://huggingface.co/fishaudio/drama-3"}],"entry":"2026-10-13-fish-audio-drama-3","notes":"Preview since 2026-09-23 (`drama-3-preview`). Weights about 39.6 GB plus a 1.87 GB codec; about 80 languages (the HF card lists 83 codes). Artificial Analysis (2026-10-13): 'Drama 3 Preview' #10 overall, Elo 1243, top open-weights model, below Eleven v4 (1328-1330). Benchmarks are vendor-run and mixed: lower error rate than ElevenLabs v4 in 9 of 24 languages, weaker in Japanese, Indonesian, Greek, Romanian. HF card metadata still names the Research License, but the LICENSE file is the Community License. No watermark is documented for the open weights.","verified":"2026-10-13","body_md":"Sources: https://huggingface.co/fishaudio/drama-3 , https://huggingface.co/fishaudio/drama-3/blob/main/LICENSE , https://github.com/fishaudio/fish-speech , https://openrouter.ai/fish-audio/drama-3 , https://artificialanalysis.ai/text-to-speech/leaderboard\n\nQuick start (Fish Audio API): `POST https://api.fish.audio/v1/tts` with header `model: drama-3`, a bearer API key, and a JSON body with `text` and a `reference_id` (voice). Self-hosting: `sgl-omni serve --model-path fishaudio/drama-3 --port 8000` (from the model card).","page_url":"https://postcutoff.com/m/fish-audio-drama-3/","standing":{"kind":"speech and audio","newer_count":0,"newer":[],"before":{"id":"fish-audio-s2","name":"Fish Audio S2 Pro / S2.1 Pro","org":"Fish Audio","released":"2026-03-09","page_url":"https://postcutoff.com/m/fish-audio-s2/"},"after":null,"weights":"public","licence":"Fish Audio Community License v1.0 (research and non-commercial free; commercial use only under USD 500,000 annual revenue; no hosted services without a separate agreement)","licence_permissive":false,"summary":"Weights are public. Licence: Fish Audio Community License v1.0. See the licence for commercial terms."}}