Text-to-speech

Fish Audio Drama 3 requirements, license & speed

By Fish Audio · released 2026-10-01 · 35B MoE (3B active per token) parameters

Commercial useConditional
Min memoryNot stated
Runs on CPUUnknown
StreamingYes

Key facts

License
Fish Audio Community License source
HF metadata tag says fish-audio-research-license; card text and LICENSE say Community License
Commercial use
Conditional source
Free commercial use only below an annual revenue threshold; hosted services need a separate license
Parameters
35B MoE (3B active per token) source
Minimum memory
Not stated by source
Runs on CPU
Unknown
Streaming
Yes source
<100 ms first audio on one H200 (vendor figure)
Voice cloning
Yes source
Languages
zh, en, ja, ko, es, pt, ar, ru, fr, de, sv, it, tr, no, nl, cy, eu, ca, da, gl, ta, hu, fi, pl, et, hi, la, ur, th, vi, jw, bn, yo, sl, cs, sw, nn, he, ms, uk, id, kk, bg, lv, my, tl, sk, ne, fa, af, el, bo, hr, ro, sn, mi, yi, am, be, km, is, az, sd, br, sq, ps, mn, ht, ml, sr, sa, te, ka, bs, pa, lt, kn, si, hy, mr, as, gu, fo source
Released
2026-10-01 source
Hugging Face repo creation date (proxy for release date)
Gated on HF
No

Measured speed & memory

Benchmark pending. We haven't measured Fish Audio Drama 3 yet. Planned targets: Raspberry Pi 5, CPU-only x86, Apple Silicon Mac and consumer GPUs. Get notified.

Accuracy

We don't run accuracy evals. For quality and intelligibility, check the Hugging Face Open TTS Leaderboard.

Similar TTS models

Spotted an error? Report it. Every fact links to the model card or repo it came from.

New open speech models: will they run on your hardware?

One email a week: new open TTS/STT models, license gotchas, and what we measured on CPUs, Macs, Pis and consumer GPUs.