Skip to content

feat(smallestai): add word_timestamps support to TTS #5921

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

Open
harshitajain165 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

feat(smallestai): add word_timestamps support to TTS #5921

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
17 changes: 17 additions & 0 deletions livekit-plugins/livekit-plugins-smallestai/livekit/plugins/smallestai/tts.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,6 +55,7 @@ class _TTSOptions:
output_format: TTSEncoding | str
base_url: str
ws_url: str
word_timestamps: bool


class TTS(tts.TTS):
Expand All @@ -70,6 +71,7 @@ def __init__(
output_format: TTSEncoding | str = "pcm",
base_url: str = SMALLEST_BASE_URL,
ws_url: str = SMALLEST_WS_URL,
word_timestamps: bool = False,
http_session: aiohttp.ClientSession | None = None,
) -> None:
"""
Expand All @@ -92,6 +94,12 @@ def __init__(
"ulaw", "alaw"). WebSocket streaming always returns PCM.
base_url: Base URL for the Smallest AI HTTP API.
ws_url: WebSocket URL for low-latency streaming synthesis.
word_timestamps: When True, the server interleaves word_timestamp events
with audio chunks on WebSocket streaming; each carries {id, word,
start, end} where start/end are seconds relative to the audio stream
start. Supported on base-queue English + Hindi voices (meher, devansh,
kartik, maithili, liam, avery). Other voice families silently emit no
word events. Defaults to False.
http_session: An existing aiohttp ClientSession to use.
"""
super().__init__(
Expand Down Expand Up @@ -120,6 +128,7 @@ def __init__(
output_format=output_format,
base_url=base_url,
ws_url=ws_url,
word_timestamps=word_timestamps,
)
self._session = http_session
self._pool = utils.ConnectionPool[aiohttp.ClientWebSocketResponse](
Expand Down Expand Up @@ -167,6 +176,7 @@ def update_options(
sample_rate: NotGivenOr[int] = NOT_GIVEN,
language: NotGivenOr[str] = NOT_GIVEN,
output_format: NotGivenOr[TTSEncoding | str] = NOT_GIVEN,
word_timestamps: NotGivenOr[bool] = NOT_GIVEN,
) -> None:
"""Update TTS options."""
if is_given(model):
Expand All @@ -181,6 +191,8 @@ def update_options(
self._opts.language = LanguageCode(language)
if is_given(output_format):
self._opts.output_format = output_format
if is_given(word_timestamps):
self._opts.word_timestamps = word_timestamps

def synthesize(
self,
Expand Down Expand Up @@ -307,6 +319,9 @@ async def _run_ws(self, text: str, output_emitter: tts.AudioEmitter) -> None:
else self._opts.language,
}

if self._opts.word_timestamps:
payload["word_timestamps"] = True

async with self._tts._pool.connection(timeout=self._conn_options.timeout) as ws:
self._acquire_time = self._tts._pool.last_acquire_time
self._connection_reused = self._tts._pool.last_connection_reused
Expand Down Expand Up @@ -337,6 +352,8 @@ async def _run_ws(self, text: str, output_emitter: tts.AudioEmitter) -> None:
audio_b64 = event.get("data", {}).get("audio")
if audio_b64:
output_emitter.push(base64.b64decode(audio_b64))
elif status == "word_timestamp":
pass # per-word timing events; no dedicated frame type in livekit-agents
elif status == "complete":
output_emitter.end_segment()
break
Expand Down
Loading