Remove extra pages (news, usage, helpwanted)

This commit is contained in:
Brandon Thomas committed 2020-07-04 02:55:24 -04:00
1 parent 2818367339
commit 8dc797819e
7 files changed
+174 -205

No files matched your search

+3 -11
View File
@@ -1,13 +1,11 @@
import './App.scss';
import React from 'react';
import { Footer } from './navigation/Footer';
import { HelpWantedComponent } from './modes/help_wanted/HelpWantedComponent';
import { Mode } from './AppMode';
import { NewsComponent } from './modes/news/NewsComponent';
import { SpeakComponent } from './modes/speak/SpeakComponent';
import { TermsComponent } from './modes/terms/TermsComponent';
import { TopNav } from './navigation/TopNav';
import { UsageComponent } from './modes/usage/UsageComponent';
import { AboutComponent } from './modes/about/AboutComponent';
import { Speaker, SPEAKERS } from './Speakers';
import { ExtrasMode } from './modes/speak/extras/ExtrasComponent';
import { Spectrogram } from './modes/speak/extras/Spectrogram';
@@ -126,14 +124,8 @@ class App extends React.Component<Props, State> {
case Mode.HISTORY_MODE:
component = <HistoryComponent utterances={this.state.utterances} />
break;
case Mode.USAGE_MODE:
component = <UsageComponent resetModeCallback={this.resetMode} />;
break;
case Mode.NEWS_MODE:
component = <NewsComponent resetModeCallback={this.resetMode} />;
break;
case Mode.HELP_WANTED_MODE:
component = <HelpWantedComponent resetModeCallback={this.resetMode} />;
case Mode.ABOUT_MODE:
component = <AboutComponent resetModeCallback={this.resetMode} />;
break;
case Mode.TERMS_MODE:
component = <TermsComponent resetModeCallback={this.resetMode} />;
+1 -3
View File
@@ -1,9 +1,7 @@
enum Mode {
SPEAK_MODE,
HISTORY_MODE,
USAGE_MODE,
NEWS_MODE,
HELP_WANTED_MODE,
ABOUT_MODE,
TERMS_MODE,
}
+169
View File
@@ -0,0 +1,169 @@
import React from 'react';
interface Props {
resetModeCallback: () => void,
}
function AboutComponent(props: Props) {
return (
<div id="usage">
<h1> Updates </h1>
<h2>
July 4, 2020: Website improvements.
</h2>
<p>
Now you can download and reply previously generated speech from the "History" tab.
I'll also work on testing the website on a variety of browsers, platforms, and
devices to ensure it works. Making the website friendly for mobile devices is also
a priority.
</p>
<h2>
June 14, 2020: New website.
</h2>
<p>
I'm creating models for a variety of speakers.
</p>
<h2>
June 7, 2020: Website rewrite.
</h2>
<p>
I rewrote the incredibly dated frontend in React and added help, news, and other
information. The frontend won't be a huge concern for me going forward, but I'll
periodically continue to make upgrades. One critical feature I intend to add is the
ability to download audio files and replay them.
</p>
<h2>
June 4, 2020: There's a new model.
</h2>
<p>
I replaced the old model with a completely new one that doesn't crash and is a
10x speed up over the last one. While it doesn't currently sound as good as the
old model, the performance improvements were good enough to warrant using it
immediately. I'll make it sound better soon. So many things in the pipeline...
</p>
<h2>May 30, 2020: Intermittent failure.</h2>
<p>
There's now another issue with how models are distributed to the workers. I have
a solution in the works and expect it to be in place before morning tomorrow.
</p>
<h2>
May 25, 2020: I'm actively working on development.
</h2>
<p>
This site has been sitting in limbo for four years. It used to use concatenative
TTS, but now uses Tacotron. I've been working on ML for the past year, but haven't
taken the time to update this website.
</p>
<p>
I've got an improved model in the works, and I'm focused on scaling the cluster.
There's an occasional server crash bug I'm chasing down (right now the server
instances respawn when they die). The frontend will be updated soon to no longer
cache results.
</p>
<h1>Usage Recommendations</h1>
<h2>Short sentences are bad</h2>
<p>
Don't say things like "hello" or "hi". They're too short for the current
model to generate good audio from. This may improve in the future, but it
is low on the list of priorities.
</p>
<h2>Try and try again</h2>
<p>
If you're unhappy with how the results sound, try changing the words a litle.
Experiment. Try adding punctuation, such as periods and question marks.
</p>
<h2>Use real words</h2>
<p>
"Asdfagdadf" is not a word, and Trump won't know how to say it.
</p>
<h2>Sound out difficult words</h2>
<p>
If the words you want to use aren't working, that's likely because I trained the
network on phonemes instead of raw text. This phonetic information
comes from a database, and although this database contains over 130,000 words, it
unfortunately doesn't have entries such as "Fortnite".
</p>
<p>
To make this work, try sounding out the problematic words: "pikachu is a pokemon"
can be rewritten as "peek ah choo is a poke ay mon". You can try "Fort Night"
instead of "Fortnite".
</p>
<p>
I'll fix this issue as soon as I can, because I know how important this is to you.
</p>
<h2>Example sentences</h2>
<p>Try getting Trump to say the following:</p>
<ul>
<li>Why does it always rain on Christmas? It's so depressing.</li>
<li>I really appreciate how good of a friend you are.</li>
<li>You're clearly not trying hard enough. What's up with you?</li>
<li>The Dinosaurs were the dominant land animals of the Mesozoic era.
Over 500 different genera of dinosaurs are known. Fossils of dinosaurs
have been found on every continent, and there are still frequent
new discoveries.</li>
<li>Dinosaurs had adaptations which helped make them successful.
The first known dinosaurs were small predators that walked on two legs.
All their descendants had an upright posture, with the legs underneath
the body. This transformed their whole lifestyle.</li>
<li>Are you really reading this far? Good for you.</li>
</ul>
<h1>Help Wanted</h1>
<h2>Training Data</h2>
<p>
Are you currently creating TTS models? I'd be happy to pay to license
your annotated audio samples. Tell me what voices you have, the sample
rate, how noisy the data is, and a little about your annotation process.
</p>
<h2>3D Modeller</h2>
<p>
I'm looking for a 3D artist or animator to make custom models and
animations for various characters (speaking, walking, articulating).
Environment, terrain, and world building skills are a huge plus.
I could rip models from VR Chat and make an attempt myself, but I'm
quite busy with working on the speech engine.
</p>
<p>
This is a paid position. It isn't necessarily for this project.
</p>
<h2>Contact</h2>
<p>Send me a message on Twitter or Gmail (same handle).</p>
<button onClick={() => props.resetModeCallback()}>Go Back</button>
</div>
)
}
export { AboutComponent };
@@ -1,41 +0,0 @@
import React from 'react';
interface Props {
resetModeCallback: () => void,
}
function HelpWantedComponent(props: Props) {
return (
<div id="help-wanted">
<h1>Help Wanted</h1>
<h2>Training Data</h2>
<p>
Are you currently creating TTS models? I'd be happy to pay to license
your annotated audio samples. Tell me what voices you have, the sample
rate, how noisy the data is, and a little about your annotation process.
</p>
<h2>3D Modeller</h2>
<p>
I'm looking for a 3D artist or animator to make custom models and
animations for various characters (speaking, walking, articulating).
Environment, terrain, and world building skills are a huge plus.
I could rip models from VR Chat and make an attempt myself, but I'm
quite busy with working on the speech engine.
</p>
<p>
This is a paid position. It isn't necessarily for this project.
</p>
<h2>Contact</h2>
<p>Send me a message on Twitter or Gmail (same handle).</p>
<button onClick={() => props.resetModeCallback()}>Go Back</button>
</div>
)
}
export { HelpWantedComponent };
-70
View File
@@ -1,70 +0,0 @@
import React from 'react';
interface Props {
resetModeCallback: () => void,
}
function NewsComponent(props: Props) {
return (
<div id="updates">
<h1> Updates </h1>
<h2>
June 14, 2020: New website.
</h2>
<p>
I'm creating models for a variety of speakers.
</p>
<h2>
June 7, 2020: Website rewrite.
</h2>
<p>
I rewrote the incredibly dated frontend in React and added help, news, and other
information. The frontend won't be a huge concern for me going forward, but I'll
periodically continue to make upgrades. One critical feature I intend to add is the
ability to download audio files and replay them.
</p>
<h2>
June 4, 2020: There's a new model.
</h2>
<p>
I replaced the old model with a completely new one that doesn't crash and is a
10x speed up over the last one. While it doesn't currently sound as good as the
old model, the performance improvements were good enough to warrant using it
immediately. I'll make it sound better soon. So many things in the pipeline...
</p>
<h2>May 30, 2020: Intermittent failure.</h2>
<p>
There's now another issue with how models are distributed to the workers. I have
a solution in the works and expect it to be in place before morning tomorrow.
</p>
<h2>
May 25, 2020: I'm actively working on development.
</h2>
<p>
This site has been sitting in limbo for four years. It used to use concatenative
TTS, but now uses Tacotron. I've been working on ML for the past year, but haven't
taken the time to update this website.
</p>
<p>
I've got an improved model in the works, and I'm focused on scaling the cluster.
There's an occasional server crash bug I'm chasing down (right now the server
instances respawn when they die). The frontend will be updated soon to no longer
cache results.
</p>
<button onClick={() => props.resetModeCallback()}>Go Back</button>
</div>
)
}
export { NewsComponent };
@@ -1,76 +0,0 @@
import React from 'react';
interface Props {
resetModeCallback: () => void,
}
function UsageComponent(props: Props) {
return (
<div id="usage">
<h1>Usage Recommendations</h1>
<h2>Short sentences are bad</h2>
<p>
Don't say things like "hello" or "hi". They're too short for the current
model to generate good audio from. This may improve in the future, but it
is low on the list of priorities.
</p>
<h2>Try and try again</h2>
<p>
If you're unhappy with how the results sound, try changing the words a litle.
Experiment. Try adding punctuation, such as periods and question marks.
</p>
<h2>Use real words</h2>
<p>
"Asdfagdadf" is not a word, and Trump won't know how to say it.
</p>
<h2>Sound out difficult words</h2>
<p>
If the words you want to use aren't working, that's likely because I trained the
network on phonemes instead of raw text. This phonetic information
comes from a database, and although this database contains over 130,000 words, it
unfortunately doesn't have entries such as "Fortnite".
</p>
<p>
To make this work, try sounding out the problematic words: "pikachu is a pokemon"
can be rewritten as "peek ah choo is a poke ay mon". You can try "Fort Night"
instead of "Fortnite".
</p>
<p>
I'll fix this issue as soon as I can, because I know how important this is to you.
</p>
<h2>Example sentences</h2>
<p>Try getting Trump to say the following:</p>
<ul>
<li>Why does it always rain on Christmas? It's so depressing.</li>
<li>I really appreciate how good of a friend you are.</li>
<li>You're clearly not trying hard enough. What's up with you?</li>
<li>The Dinosaurs were the dominant land animals of the Mesozoic era.
Over 500 different genera of dinosaurs are known. Fossils of dinosaurs
have been found on every continent, and there are still frequent
new discoveries.</li>
<li>Dinosaurs had adaptations which helped make them successful.
The first known dinosaurs were small predators that walked on two legs.
All their descendants had an upright posture, with the legs underneath
the body. This transformed their whole lifestyle.</li>
<li>Are you really reading this far? Good for you.</li>
</ul>
<button onClick={() => props.resetModeCallback()}>Go Back</button>
</div>
)
}
export { UsageComponent };
+1 -4
View File
@@ -11,10 +11,7 @@ function TopNav(props: Props) {
<nav>
<a href="#speak" onClick={() => props.switchModeCallback(Mode.SPEAK_MODE)}>Speak</a>
<a href="#history" onClick={() => props.switchModeCallback(Mode.HISTORY_MODE)}>History</a>
{/*<a onClick={() => props.switchModeCallback(Mode.HISTORY_MODE)}>Downloads</a>*/}
<a href="#use" onClick={() => props.switchModeCallback(Mode.USAGE_MODE)}>Usage</a>
<a href="#news" onClick={() => props.switchModeCallback(Mode.NEWS_MODE)}>News</a>
<a href="#help" onClick={() => props.switchModeCallback(Mode.HELP_WANTED_MODE)}>Help Wanted</a>
<a href="#use" onClick={() => props.switchModeCallback(Mode.ABOUT_MODE)}>About</a>
<hr />
</nav>
)