Skip to content

(inference): add fish audio tts #6529

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
tinalenguyen merged 1 commit into from
Jul 23, 2026
Merged

(inference): add fish audio tts #6529

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
51 changes: 50 additions & 1 deletion livekit-agents/livekit/agents/inference/tts.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,9 +67,21 @@
"inworld/inworld-tts-1",
]
XaiModels = Literal["xai/tts-1",]
FishAudioModels = Literal[
"fishaudio",
"fishaudio/s2.1-pro",
"fishaudio/s2.1-pro-free",
"fishaudio/s2-pro",
]

TTSModels = (
CartesiaModels | DeepgramModels | ElevenlabsModels | RimeModels | InworldModels | XaiModels
CartesiaModels
| DeepgramModels
| ElevenlabsModels
| RimeModels
| InworldModels
| XaiModels
| FishAudioModels
)


Expand Down Expand Up @@ -194,6 +206,23 @@ class XaiOptions(TypedDict, total=False):
optimize_streaming_latency: Literal[0, 1, 2] # latency optimization level, default 0


class FishAudioOptions(TypedDict, total=False):
"""See https://docs.fish.audio/api-reference/endpoint/websocket/tts-live"""

speed: float # prosody speaking-rate multiplier, 1.0 is natural
volume: float # prosody loudness adjustment in dB, 0 is natural
normalize_loudness: bool # consistent output loudness; S2-Pro family only
temperature: float # range 0-1, higher is more varied/expressive
top_p: float # nucleus sampling probability mass, range 0-1
normalize: bool # normalize numbers/dates/abbreviations before synthesis
latency: Literal["normal", "balanced", "low"] # streaming latency mode
chunk_length: int # characters buffered before auto-synthesis, range 100-300
max_new_tokens: int # max audio tokens per text chunk, default 1024
min_chunk_length: int # min characters before splitting a new chunk, range 0-100
condition_on_previous_chunks: bool # use prior audio as context for consistency
early_stop_threshold: float # early-stopping threshold for batching, range 0-1


TTSEncoding = Literal["pcm_s16le"]

DEFAULT_ENCODING: TTSEncoding = "pcm_s16le"
Expand Down Expand Up @@ -330,6 +359,25 @@ def __init__(
) -> None:
pass

@overload
def __init__(
self,
model: FishAudioModels,
*,
voice: NotGivenOr[str] = NOT_GIVEN,
language: NotGivenOr[str] = NOT_GIVEN,
encoding: NotGivenOr[TTSEncoding] = NOT_GIVEN,
sample_rate: NotGivenOr[int] = NOT_GIVEN,
base_url: NotGivenOr[str] = NOT_GIVEN,
api_key: NotGivenOr[str] = NOT_GIVEN,
api_secret: NotGivenOr[str] = NOT_GIVEN,
http_session: aiohttp.ClientSession | None = None,
extra_kwargs: NotGivenOr[FishAudioOptions] = NOT_GIVEN,
fallback: NotGivenOr[list[FallbackModelType] | FallbackModelType] = NOT_GIVEN,
conn_options: NotGivenOr[APIConnectOptions] = NOT_GIVEN,
) -> None:
pass

@overload
def __init__(
self,
Expand Down Expand Up @@ -369,6 +417,7 @@ def __init__(
| RimeOptions
| InworldOptions
| XaiOptions
| FishAudioOptions
] = NOT_GIVEN,
fallback: NotGivenOr[list[FallbackModelType] | FallbackModelType] = NOT_GIVEN,
conn_options: NotGivenOr[APIConnectOptions] = NOT_GIVEN,
Expand Down
Loading