[Voxtral] Fix speech transcription api (#31388)

Signed-off-by: Patrick von Platen <patrick.v.platen@gmail.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Signed-off-by: bk-201 <joy25810@foxmail.com>
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
Signed-off-by: prashanth058 <prashanth.dannamaneni@uipath.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
Co-authored-by: bk-201 <joy25810@foxmail.com>
Co-authored-by: prashanth058 <prashanth.dannamaneni@uipath.com>
Co-authored-by: Anexdeus <5142168@mail.ru>
Co-authored-by: Julien Denize <40604584+juliendenize@users.noreply.github.com>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
Co-authored-by: Nicolò Lucchesi <nicolo.lucchesi@gmail.com>
This commit is contained in:
Patrick von Platen
2026-01-08 18:34:19 +08:00
committed by GitHub
co-authored by Harry Mellor Jee Jee Li bk-201 prashanth058 Anexdeus Julien Denize Cyrus Leung Nicolò Lucchesi
parent 2972a05473
commit 18d4e481d0
5 changed files with 114 additions and 27 deletions
+13 -1
View File
@@ -477,7 +477,15 @@ class OpenAISpeechToText(OpenAIServing):
}
segment_class: type[SpeechToTextSegment] = segments_types[self.task_type]
text = ""
chunk_size_in_s = self.asr_config.max_audio_clip_s
if chunk_size_in_s is None:
assert len(list_result_generator) == 1, (
"`max_audio_clip_s` is set to None, audio cannot be chunked"
)
for idx, result_generator in enumerate(list_result_generator):
start_time = (
float(idx * chunk_size_in_s) if chunk_size_in_s is not None else 0.0
)
async for op in result_generator:
if request.response_format == "verbose_json":
segments: list[SpeechToTextSegment] = (
@@ -485,7 +493,7 @@ class OpenAISpeechToText(OpenAIServing):
tokens=tuple(op.outputs[0].token_ids),
segment_class=segment_class,
request=request,
start_time=idx * self.asr_config.max_audio_clip_s,
start_time=start_time,
)
)
@@ -653,6 +661,10 @@ class OpenAISpeechToText(OpenAIServing):
def _split_audio(
self, audio_data: np.ndarray, sample_rate: int
) -> list[np.ndarray]:
assert self.asr_config.max_audio_clip_s is not None, (
f"{self.asr_config.max_audio_clip_s=} cannot be None to"
" split audio into chunks."
)
chunk_size = sample_rate * self.asr_config.max_audio_clip_s
overlap_size = sample_rate * self.asr_config.overlap_chunk_second
chunks = []