TaSTT-Whisper.git/Whisper/MF, branch masterHigh-performance GPGPU inference of OpenAI's Whisper automatic speech recognition (ASR) modelhttps://git.yummers.dev/TaSTT-Whisper.git/atom?h=master2023-04-05T00:41:30+00:00Fix audio normalization2023-04-05T00:41:30+00:00yumyum.food.vr@gmail.com2023-04-05T00:40:59+00:00urn:sha1:59297502afb8f61c1216c6d57d6cc18ab5b9f467
Normalization was putting audio onto range [0, 255], while it should
have been on range [0, 1].
* Add AudioBuffer::save() to enable debugging audio issues.
Begin work on evaluation framework2023-03-03T23:43:24+00:00yumyum.food.vr@gmail.com2023-03-03T23:43:24+00:00urn:sha1:f7d5741e5c069d759f8412bd40b279e1d7abac4c
Need a way to verify that beam search is actually working better than
greedy.
Add retainDuration option to CaptureParams2023-02-27T04:09:15+00:00yumyum.food.vr@gmail.com2023-02-27T03:42:33+00:00urn:sha1:1136acfc365f357d2df13a263714e8ae0614c4f9
This allows users to retain a suffix of the PCM buffer after a VAD
segmentation event, reducing some instances of words being lost at
the start of the next VAD window.
Normalize audio before sending to transcription layer2023-02-27T04:08:45+00:00yumyum.food.vr@gmail.com2023-02-27T03:12:01+00:00urn:sha1:02c2605454288f7c86023ae700366acf08cd2206
Helps in cases where the speaker is speaking softly, or their mic gain
is set low.
Frames with no VAD are shortened, not dropped2023-02-27T03:49:40+00:00yumyum.food.vr@gmail.com2023-02-27T02:57:07+00:00urn:sha1:00a0350a0218cf4b03d14acac84110bc1e882bee
On PCM buffers of length >= captureParams.dropStartSilence, a
"no voice" VAD verdict would result in the PCM buffer being entirely
cleared. The emergent behavior is that when VAD segments speech,
words right after the segmentation window can frequently be dropped.
By removing a prefix from the PCM buffer and clearing the VAD buffers,
the transcription algorithm has access to "leading" frames before the
frames which triggered VAD. This reduces cases where words are omitted
in the middle of long statements.
Bugfix, stereo PCM handling2023-01-28T15:15:39+00:00Konstantinconst@const.me2023-01-28T15:15:39+00:00urn:sha1:cfd20a0f796ab6cc046b080bb7af8967cb7c361bMinor, error handling2023-01-20T12:31:49+00:00Konstantinconst@const.me2023-01-20T12:31:49+00:00urn:sha1:8880867a950f78292f8cf4b37771b08dd0376074Workaround for the Microsoft’s bug in their MP3 decoder MFT2023-01-19T16:10:24+00:00Konstantinconst@const.me2023-01-19T16:10:24+00:00urn:sha1:9df2ee2ead4ce23d06351a6cdb4fea588f79e429Minor, optimized away memcpy() when running audio capture2023-01-17T12:12:14+00:00Konstantinconst@const.me2023-01-17T12:12:14+00:00urn:sha1:aae8860d0e7b2bf68a1c20c6f30999fff531f03cSource codes2023-01-16T13:52:43+00:00Konstantinconst@const.me2023-01-16T13:52:43+00:00urn:sha1:8c4603c73675958efc960fbd4bb599a2909d106a