wx44wx/pipecat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

servceis(cartesia,elevenlabs): use word start times instead

Browse Source
This commit is contained in:
Aleix Conchillo Flaqué
2024-09-13 13:10:44 -07:00
parent cb36a71381
commit 1fe940bd6b
2 changed files with 6 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/pipecat/services/cartesia.py
View File

@@ -177,7 +177,7 @@ class CartesiaTTSService(AsyncWordTTSService):
await self.add_word_timestamps([("LLMFullResponseEndFrame", 0)]) await self.add_word_timestamps([("LLMFullResponseEndFrame", 0)])
elif msg["type"] == "timestamps": elif msg["type"] == "timestamps":
await self.add_word_timestamps( await self.add_word_timestamps(
list(zip(msg["word_timestamps"]["words"], msg["word_timestamps"]["end"])) list(zip(msg["word_timestamps"]["words"], msg["word_timestamps"]["start"]))
) )
elif msg["type"] == "chunk": elif msg["type"] == "chunk":
await self.stop_ttfb_metrics() await self.stop_ttfb_metrics()

15
src/pipecat/services/elevenlabs.py
View File

@@ -51,22 +51,17 @@ def sample_rate_from_output_format(output_format: str) -> int:
def calculate_word_times( def calculate_word_times(
alignment_info: Mapping[str, Any], cumulative_time: float alignment_info: Mapping[str, Any], cumulative_time: float
) -> List[Tuple[str, float]]: ) -> List[Tuple[str, float]]:
end_times = [ zipped_times = list(zip(alignment_info["chars"], alignment_info["charStartTimesMs"]))
s + d for s,
d in zip(
alignment_info["charStartTimesMs"],
alignment_info["charDurationsMs"])]
zipped_end_times = list(zip(alignment_info["chars"], end_times))
words = "".join(alignment_info["chars"]).split(" ") words = "".join(alignment_info["chars"]).split(" ")
# Calculate end time for each word. We do this by finding a space character # Calculate start time for each word. We do this by finding a space character
# and using the previous word time, also taking into account there might not # and using the previous word time, also taking into account there might not
# be a space at the end. # be a space at the end.
times = [] times = []
for (i, (a, b)) in enumerate(zipped_end_times): for (i, (a, b)) in enumerate(zipped_times):
if a == " " or i == len(zipped_end_times) - 1: if a == " " or i == len(zipped_times) - 1:
t = cumulative_time + (zipped_end_times[i - 1][1] / 1000.0) t = cumulative_time + (zipped_times[i - 1][1] / 1000.0)
times.append(t) times.append(t)
word_times = list(zip(words, times)) word_times = list(zip(words, times))