SpeechPulse Version 10 Demo: Live Dictation, File Mode, and System Audio Mode
SpeechPulse Version 10 Demo: Live Dictation, File Mode, and System Audio Mode
0:00 Live Mode
16:53 File Mode
25:44 System Audio Mode
Closed Caption:
In this video, I'm going to demonstrate
some of the basic functions of speech
paths new paragraph. Right now, I'm
dictating to Windows Notepad using auto
punctuation mode.
I'm also running speech pulse on an
Nvidia GPU new paragraph.
As you can see right now I'm using the
realtime processing mode. In this mode,
speech pulse inserts the text as you
dictate. New paragraph.
The speech pulse also has few other
dictation modes where you can control
how it works. New paragraph.
Now I'm going to switch off the real
time processing mode and switch to the
automatic speech input mode. Feel
better, bro.
Now speech passes using automatic speech
input mode with auto spacing.
That means speech p only inserts the
text after you complete the dictation.
With auto spacing mode, speech pass
scans the previous character in the text
edit to determine if it should enter a
space character before the next speech
segment.
Now I'm going to switch speech pulse to
the manual spacing mode. With manual
spacing, speech pulse doesn't scan the
previous character and immediately
inserts the text. It is faster, but you
have to enter the space character
manually. New paragraph.
You can also minimize the speech pulse
window or reduce it to the taskbar new
paragraph.
The speech pulse also supports pushto
talk input. Now I'm going to switch
speech files to the push to talk mode.
New paragraph.
There are two pushto talk modes and
right now I'm going to use the push to
talk press and hold mode.
Right now I'm using the push to talk
press and hold mode. With this mode, you
have to press and hold the pushto talk
hot key to start the
transcription and then release the hot
key to trigger the processing.
There are various push to talk hot keys.
You can find all of them and configure
them in the options window.
This is the primary pushto talk hotkey.
And there are also various other push to
do hot keys.
The speech pulse also supports manual
punctuation.
at the moment only for the English and
German languages. New paragraph. Now I'm
going to switch speech pass to the
manual punctuation mode. New paragraph.
Now I'm using the manual punctuation
mode. Period. With manual punctuation
mode, you need to manually dictate the
punctuation marks. Period. New
paragraph. Now I'm going to switch back
to the auto punctuation mode. Period.
New paragraph.
Speech pass also has a history window
where you can find your previous
dictations.
You can also retranscribe any of your
previous dictations using different
speech models, language models and other
settings using the history window.
The speech pulse also has a history
window.
The speech pulse also has a setting
where you can add your custom mappings
to correct any inaccuracies in your
dictation. I have already added
one mapping for the
word speech pulse.
Here speech pulse will replace these
combinations with the correct
version. You can add more mappings here
using this button. Speech
P also supports
postprocessing your dictations using
large language models. New paragraph.
Now I'm going to use the standard
English language model which you can
download and demonstrate how it works.
New paragraph.
You can download language models using
the language model download.
Currently, it only has English standard
model which I have already downloaded.
I'm also going to switch the language
model to the GPU.
Similarly, you can also add additional
speech models using the speech model
download.
There are various models here and larger
models have better accuracy but smaller
models run faster.
Now I'm running both these models on the
GPU, Nvidia GPU. You can also run both
of them on the CPU, but the CPU
processing will be slower than the GPU.
Now you need to select an AI template to
process your dictation. English standard
model has few default templates you can
use.
Now I'm going to switch use the email
example template and see how it will
work. You can edit these templates
using this edit button. Default
templates don't support editing but you
can add your custom templates and edit
them here.
In this default template,
we are instructing the AI language model
to format the input in a format suitable
for email communication.
When this email template is active,
you can dictate something like this. And
it the output will be formatted like
this.
I can't come to your party today,
Michael. I have a meeting to attend.
As you can see, the AI language model
correctly formats your dictation
according to your email template.
You can also add various speech and LLM
APIs to speech files.
You can add speech APIs using the add
more APIs button from the speech model
dropd down.
Here you can enter the API name and the
base URL and then the model name and
your API key.
Similarly, you can add LLM or large
language model APIs using the add more
APIs
option.
Here you can enter the API name, then
base URL, then API key.
For example, now I'm going to use the
open AI API and select one of the
models.
Here you have various models from the
open AI which you can use. You can also
search them from here.
As you can see, you can select one of
them and then connect.
Okay. Now we are using this DPT 4.1 nano
API as our language model.
Then you can add a template
based on one of the available templates.
You can edit them
and
pay what you want from this
API.
You can also add examples example input
output combination so the uh AI model
can understand what you your actual
requirement.
Now I'm using the chat API so you can
ask anything and
they I will give you the output.
What is the distance between earth and
sun?
This is from the open AI nano model from
the API.
You can also train new words to speech
pulse using the training option.
First you need to enable training and
enable GPU acceleration. If you have a
GPU then you need to restart speech
pass. I have already done those. So
I'm going to continue.
This training window has option to add
training profiles. When you add training
profiles, you can name them and then
select one of the supported languages.
Currently supports these languages.
You can then you have to select that
training profile. Then you can add new
word using this window.
You can also add a different
pronunciation if you have one here.
Benica HCT.
Benica HCT.
Benica HCT.
Benica HCT.
Benica HC.
Benica XCT.
Benica XCT.
Benica XCT.
Now I'm going to dictate a sentence
with the trend word and see how it
works.
Penica HCT is a fixed dose combination
medication used to treat high blood
pressure.
As you can see, speech pulse now
correctly recognize this word.
Now I'm going to switch speech pulse to
the file mode.
File mode supports batch file
transcription and subtitle generation.
I'm also going to disable AI processing.
You can also process your file
transcription outputs using AI models.
For example, to generate summaries of
your audio files. But for this
demonstration, I'm going to disable AI
processing. Now I'm using the multilarge
model on the GPU
for English transcription.
Now I'm going to transcribe this file
using the multilash model on the GPU.
You also need to select the output
folder using this button.
This is the transcription output window.
This window supports various shortcuts
to play back the audio. For example, you
can simply click on a word to start
playback from that word. You can also
use the tap key for play pause.
Hello, this is Graeme Freyer. I'm in an
office in Windgate Road composites with
me is very much so we actually started
and yet you hardly made a dent on this
premises. So about 35 people in this we
actually have a background with uh need
to take into consideration
whether it's going to be too hot. The
range of possibilities for us is very
hard to find local economy that you
think will help us grow a little bit
quicker and faster. Sounds like a good
future to me.
You can also save your transcriptions
for future editing using this save
transcription button.
Hello, this is Grant. You hardly made a
death.
Transcription window also has find and
replace functionality.
You can also retranscribe any of the
incorrectly transcribed sections.
Simply select the segment and then right
click
retanscribe selection.
very much. So we actually start you
hardly made a dent on this premises.
This window supports all types of
editing. So you can correct your
dictation,
correct your transcription directly in a
bit of a rapid and you hardly made it.
got a twotory office twotory office
block and there's currently three of us
in here. Um so our plans up to about 35
people in this according to plan would
need to upscale again in 3 to four years
time. Wow, that's some some dramatic
goes according to plan would need to
upscale.
You can also export your final
transcription using various output
formats.
This subtitle format and this is also
another subtitle format,
paragraphs,
sentences.
Then this also paragraphs, a full edited
text.
You can also generate per word
synchronized uh subtitles by setting
word count per line or character count
per line.
As you can see now the subtitles have
one word per line. So each word is
synchronized when you play back on the
editor.
Speech
pass also supports
speaker automatic speaker diorization.
There are two methods. The recommended
one is the primary method.
Now I'm going to
perform speaker diation on the same
audio file.
This is the diorization output window.
As you can see, speech pulse are
separated speakers. There are two
speakers automatically detected.
You can rename these speakers using edit
speaker names button.
When you replace speakers here, speech
pass saves the speaker identification
data. So the next time these names will
be automatically filled
for I mean and taken over by very much
so we actually started in a 1,500 ft uh
barn in Sway Hampshire we managed you've
hardly made a dent on this premises so I
can give you a very brief blade to a
company
This window also supports
retranscription.
Yeah. So, we're still in here. And then
if the project three
You can also edit text here.
And if you
if a single segment has two speakers
incorrectly detected, you can split the
segments by pressing the enter key.
Then you can name it
from here.
Similarly, if a single speaker is
separated to two segments, you can press
the backspace button at the beginning of
a segment to merge it with the previous
segment.
Workshop in uh 6 months. uh we've gone
for
you can also save direct sessions using
this button
and load them from here. Hello, this is
Gra. So we actually started and then
you can find the list of speaker names
that speech pass can automatically
detect from option window.
These are the saved speakers.
When speech pass detect these speakers,
you don't have to rename them manually.
Speech pass will automatically add the
names.
Now I'm going to demonstrate the system
audio mode.
You can use system audio mode to
transcribe any audio you play back on
your computer. For example, you can
transcribe
any online videos you are watching or
tutorials
in real time using system audio mode.
System audio mode also supports
transcribing both system audio and
microphone at the same time. For
example, you can use this feature to
transcribe your online interviews. Now
I'm going to demonstrate the system
audio mode recording only the system
audio.
You can configure the system audio mode
using the options window.
You can select the source device from
here.
These are your output speakers and you
can select one of them and
set it at as the source for the system
audio.
This is only for the Windows version.
Mac OS version doesn't need this option.
Hello, this is Graeme Freyer. I'm in an
office in Windgate Road, Gosport, which
used to be the the headquarters for
North Sales, I believe, and now is taken
over by AC Marine and Composites. With
me is Alex Newton, who's the boss here.
Alex, this is an enormous factory, but
you've got enormous plans, haven't you?
Very much so. We actually started in a
1,500 ft² uh barn in Sway, Hampshire. Uh
we've managed to upscale to a 22,000t
workshop in uh 6 months. Uh we've grown
from two full-time employees to uh I
think we're on nine now in in the space
of 6 months. So, it's been a bit of a
rapid exp. You hardly made a dent on
this premises. So you've got you've got
big plans, haven't you? Yeah. So we're
we're still rattling around in the
workshop. There's uh we've got a
two-story office block and there's
currently three of us in here. Um so our
plans at the moment are to expand up to
about 35 people in this workshop. Uh and
then if the project that we're working
on goes according to plan, we'd need to
upscale again in 3 to four years time.
Wow. That's some some dramatic plan
that. So what is this project? Can you
talk about it? I can give you a very
System audio mode saves the audio and
text transcription
in your output folder. Here we have the
system transcription
and the recorded audio.
Hello, this is Gam workshop. There's uh
we've got
now I'm going to enable both system and
microphone and
transcribe them together.
Hello, this is Graeme Freyer. I'm in an
office in Windgate Road, Gosport, which
used to be the the headquarters for
North Sales, I believe. And now is taken
over by AC Marine and Composites. With
me is Alex Newton, who's the boss here.
Alex, this is an enormous factory, but
you've got enormous plans, haven't you?
Very much so. So, we actually started in
a 1,500 ft² uh barn in Sway, Hampshire.
Uh we've managed to upscale to a 22,000
foot workshop in uh 6 months. Uh we've
grown from two full-time employees to uh
I think we're on nine now in in the
space of 6 months. So, it's been a bit
of a rapid exp. And you've hardly made a
dent on this premises. So, you've got
you've got big plans, haven't you? Yeah.
So, we're we're still rattling around in
the workshop.
Now we have both microphone and system
transcriptions and their recordings.
North sales I believe
I believe you also have the combined
transcription.
There are various text formats you can
select.
Now I'm using the paragraphs with
timestamps. You can select sentences,
sentences with timestamps and
paragraphs. Also full text
transcription.
System audio mode also supports
speaker diorization.
You can enable speaker diorization
from here.
You can select active speakers
otherwise you can
add new speakers
and rename them directly in the system
mode you want.
Hello, this is Graeme Fer. I'm in an
office in Windgate Road, Gossport, which
used to be the the headquarters for
North Sales, I believe, and now is taken
over by AC Marine and Composites. With
me is Alex Newton, who's the boss here,
Alex, this is an enormous factory, but
you've got enormous plans, haven't you?
Very much so. We actually started in a
1,500 ft uh barn in Sway, Hampshire. Uh
we've managed to upscale to a 22,000
square foot workshop in uh six months.
Uh we've grown from two full-time
employees to uh I think we're on nine
now in in the space of 6 months. So it's
been a bit of a rapid experience. And
yet you hardly made a dent on this
premises. So you've got you've got big
plans, haven't you? Yeah. So we're we're
still rattling around in the workshop.
There's uh we've got a twotory office
block and there's currently three of us
in here. Um, so our plans at the moment
are to expand up to about 35 people in
this workshop. Uh, and then if the
project that we're working on goes
according to plan, we'd need to upscale
again in 3 to four years time. Wow.
That's some some some dramatic plan
that. So what is this project? Can you
talk about it? I can give you a very
brief overview of what we're doing.
Thank you for watching this
demonstration.
Video Length: 32:55
Uploaded By: SpeechPulse
Published: 6/13/2025
View Count: 208,681