dev-tts page (#627)

* tts revamp init

* revamp 2 init

* update

* removed urls voice details
This commit is contained in:
BFlat authored and GitHub committed 2024-07-24 23:34:28 +07:00
1 parent 4691647186
commit 75e1096961
10 files changed
+558 -146

No files matched your search

@@ -22,7 +22,7 @@
&.badge-small {
font-size: 11px;
padding: 1px 6px;
padding: 2px 6px;
border-radius: 0.3rem;
}
}
@@ -14,14 +14,16 @@ export default function Label({
required,
disabled,
}: Props) {
return label ? <div {...{ className: `fy-input-label${disabled ? "opacity-50" : "" }` }}>
<label {...{ className: `fw-medium${required ? " required" : "" }` }} >
{label}
</label>
{ invalidReason ? (
<span {...{ className: "label-error form-text red is-danger" }}>
{invalidReason}
</span>
) : null }
</div> : null;
return label ? (
<div {...{ className: `fy-input-label${disabled ? "opacity-50" : ""}` }}>
<label {...{ className: `fw-medium${required ? " required" : ""}` }}>
{label}
</label>
{invalidReason ? (
<span {...{ className: "label-error form-text red is-danger" }}>
{invalidReason}
</span>
) : null}
</div>
) : null;
}
@@ -8,7 +8,8 @@
position: relative;
background-color: #57577d;
img, .cover-img-edit {
img,
.cover-img-edit {
position: absolute;
width: 100%;
height: 100%;
@@ -23,32 +24,36 @@
}
.cover-img-edit {
padding: .25rem;
font-size: .75rem;
padding: 0.25rem;
font-size: 0.75rem;
cursor: pointer;
display: flex;
align-items: center;
justify-content: center;
transition: all .25s ease-in-out;
transition: all 0.25s ease-in-out;
color: transparent;
span { text-align: center }
span {
text-align: center;
}
svg {
position: absolute;
color: #fff;
transition: background .25s ease-in-out;
background: rgba(#000, .5);
border-radius: .25rem;
padding: .5rem;
right: .25rem;
bottom: .25rem;
transition: background 0.25s ease-in-out;
background: rgba(#000, 0.5);
border-radius: 0.25rem;
padding: 0.5rem;
right: 0.25rem;
bottom: 0.25rem;
}
&:hover {
color: #fff;
background: rgba(#000, .5);
svg { background: none }
background: rgba(#000, 0.5);
svg {
background: none;
}
}
}
}
@@ -68,6 +68,7 @@ export default function TopNav({
"/creator-onboarding"
);
const isOnWelcomePage = window.location.pathname === "/welcome";
const isOnTtsPage = window.location.pathname.includes("/dev-tts");
const { open } = useModal();
const openModal = () => open({ component: InferenceJobsModal });
@@ -481,11 +482,13 @@ export default function TopNav({
{(!isOnLandingPage &&
!isOnLoginPage &&
!isOnSignUpPage &&
!isOnStudioPage) ||
!isOnStudioPage &&
!isOnTtsPage) ||
(loggedIn &&
!isOnLoginPage &&
!isOnSignUpPage &&
!isOnStudioPage) ||
!isOnStudioPage &&
!isOnTtsPage) ||
(isOnLandingPage &&
isScrolled &&
!isOnLoginPage &&
@@ -282,6 +282,9 @@ styles with important tags:
.alert {
border-radius: 0.5rem !important;
padding: 0.75rem 1rem;
font-size: 14px;
font-weight: 500;
}
.alert-pricing {
@@ -1812,7 +1815,7 @@ textarea {
}
.session-tts-section {
height: 450px;
height: 370px;
}
.session-vc-section {
@@ -727,8 +727,20 @@ class PageContainer extends React.Component<
<DevMediaInput />
</Route>
<Route path="/dev-tts">
<DevTTS />
<Route path="/dev-tts/:weight_token">
<DevTTS
sessionSubscriptionsWrapper={
this.props.sessionSubscriptionsWrapper
}
/>
</Route>
<Route exact path="/dev-tts">
<DevTTS
sessionSubscriptionsWrapper={
this.props.sessionSubscriptionsWrapper
}
/>
</Route>
{/* Route for initial voice creation */}
@@ -3,11 +3,7 @@ import { Link } from "react-router-dom";
import { BucketConfig } from "@storyteller/components/src/api/BucketConfig";
import { JobState } from "@storyteller/components/src/jobs/JobStates";
import { FontAwesomeIcon } from "@fortawesome/react-fontawesome";
import {
faClock,
faLink,
faHeadphonesSimple,
} from "@fortawesome/free-solid-svg-icons";
import { faClock, faHeadphonesSimple } from "@fortawesome/free-solid-svg-icons";
import { SessionSubscriptionsWrapper } from "@storyteller/components/src/session/SessionSubscriptionsWrapper";
import { Analytics } from "../../../common/Analytics";
@@ -23,6 +19,11 @@ import {
InferenceJob,
} from "@storyteller/components/src/jobs/InferenceJob";
import { useInferenceJobs, useLocalize } from "hooks";
import { Button } from "components/common";
import {
faArrowDownToLine,
faArrowRight,
} from "@fortawesome/pro-solid-svg-icons";
interface Props {
sessionSubscriptionsWrapper: SessionSubscriptionsWrapper;
@@ -115,11 +116,7 @@ function SessionTtsInferenceResultList(props: Props) {
results.push(
<div key={job.jobToken}>
<div>
<div>
<div className={cssStyle}>{stateDescription}</div>
</div>
</div>
<div className={cssStyle}>{stateDescription}</div>
</div>
);
} else {
@@ -145,27 +142,41 @@ function SessionTtsInferenceResultList(props: Props) {
<button className="delete" aria-label="delete"></button>
</div>*/}
<div>
<div className="panel panel-tts-results p-4 gap-3 d-flex flex-column">
<div className="panel panel-tts-results p-3 gap-3 d-flex flex-column">
<div>
<h5 className="mb-2">{job.maybeModelTitle}</h5>
<p>{job.maybeRawInferenceText}</p>
<div className="d-flex align-items-center gap-1">
<h6 className="mb-1 fw-semibold flex-grow-1">
{job.maybeModelTitle}
</h6>
<Button
iconFlip={true}
variant="link"
label="More details"
className="fs-7"
icon={faArrowRight}
to={ttsPermalink}
/>
</div>
<p className="fs-7">{job.maybeRawInferenceText}</p>
</div>
{/* <audio
className="w-100"
controls
src={audioLink}
onClick={() => {
Analytics.ttsClickResultInlinePlay();
}}
>
Your browser does not support the
<code>audio</code> element.
</audio> */}
<div className="d-flex gap-3 align-items-center">
{wavesurfers}
<Button
variant="action"
small={true}
square={true}
icon={faArrowDownToLine}
fontLarge={true}
to={ttsPermalink}
onClick={() => {
Analytics.ttsClickResultLink();
}}
/>
</div>
{wavesurfers}
<div className="mt-2">
{/* <div className="mt-2">
<Link
to={ttsPermalink}
onClick={() => {
@@ -176,7 +187,7 @@ function SessionTtsInferenceResultList(props: Props) {
<FontAwesomeIcon icon={faLink} className="me-2" />
{t("resultsAudioShareDownload")}
</Link>
</div>
</div> */}
</div>
</div>
</div>
@@ -227,13 +238,9 @@ function SessionTtsInferenceResultList(props: Props) {
}
return (
<div>
<div>
<div className="d-flex flex-column gap-3">
{upgradeNotice}
<div className="d-flex flex-column gap-3">{results}</div>
</div>
</div>
<div className="d-flex flex-column gap-3">
{upgradeNotice}
<div className="d-flex flex-column gap-3">{results}</div>
</div>
);
}
@@ -1,37 +1,81 @@
@import "scss/colours.scss";
@import "scss/site-refresh.scss";
.fy-featured-voices-section {
background-color: rgba($action, 0.25);
border-radius: 0.5rem;
padding: 20px;
}
.fy-featured-voices {
background-color: $secondary-light;
border-radius: 0.5rem;
padding: 0.6rem;
border: 2px solid $action;
cursor: pointer;
transition: all 0.2s;
user-select: none;
height: 100%;
&:hover {
border-color: $primary;
background-color: rgba($action, 0.7);
}
.fy-select-voice {
color: $white;
padding: 1px 8px;
background-color: lighten($action, 5%);
border-radius: 0.35rem;
font-size: 13px !important;
}
}
.fy-weight-picker {
@include panel;
border-radius: .5rem;
background-color: rgba($secondary, .6);
border-radius: 0.5rem;
background-color: rgba($secondary, 0.6);
&.fy-weight-picker-full {
.fy-input input { border-radius: .5rem .5rem 0 0 }
}
.fy-card-badge {
display: inline-flex;
margin-left: 0.5rem;
}
.fy-weight-picker-preview {
border-top: 1px solid rgba(#fff,.15);
display: flex;
}
.fy-input {
margin-bottom: 0;
}
}
.fy-card-badge {
display: inline-flex;
margin-left: .5rem;
}
.fy-weight-picker-preview {
border-radius: 0 0 0.5rem 0.5rem;
padding: 0.6rem;
background-color: rgba($action, 0.8);
border: 2px solid lighten($action, 12%);
display: flex;
align-items: center;
transition: all 0.2s;
cursor: pointer;
.fy-input { margin-bottom: 0 }
&:hover {
background-color: rgba($action, 1);
border: 2px solid $primary;
}
}
.fy-weight-picker-reselect {
background-color: $secondary-light;
border-radius: .5rem;
padding: .5rem;
margin: 0 0 auto auto;
}
.fy-weight-picker-reselect {
background-color: $secondary;
border-radius: 0.5rem;
padding: 0.5rem;
margin: 0 0 auto auto;
transition: all 0.2s;
cursor: pointer;
&:hover {
background-color: $action;
}
}
.fy-card-badge {
display: flex;
height: min-content;
}
display: flex;
height: min-content;
}
@@ -1,39 +1,84 @@
import React, { useState } from "react";
import { Link } from "react-router-dom";
import PageHeader from 'components/layout/PageHeader';
import { MediaBrowser } from "components/modals";
import { CardBadge } from "components/entities";
import { Button, Container, Label, Panel, TempTextArea as TextArea, TempInput as Input, WeightCoverImage } from "components/common";
import { useDebounce, useModal } from "hooks";
import {
Button,
Container,
Input,
Label,
Panel,
TextArea,
WeightCoverImage,
} from "components/common";
import { useDebounce, useInferenceJobs, useModal } from "hooks";
import { BucketConfig } from "@storyteller/components/src/api/BucketConfig";
import { Weight } from "@storyteller/components/src/api/weights/GetWeight";
import { FontAwesomeIcon as Icon } from "@fortawesome/react-fontawesome";
import { faRotateLeft, faSearch } from "@fortawesome/pro-solid-svg-icons";
import { FontAwesomeIcon } from "@fortawesome/react-fontawesome";
import {
faArrowRight,
faBell,
faChevronRight,
faDeleteLeft,
faSearch,
faWaveformLines,
faXmark,
} from "@fortawesome/pro-solid-svg-icons";
import "./DevTTS.scss";
import { FeaturedVoice } from "./FeaturedVoice";
import { SessionTtsInferenceResultList } from "v2/view/_common/SessionTtsInferenceResultsList";
import StorytellerStudioCTA from "components/common/StorytellerStudioCTA";
import { usePrefixedDocumentTitle } from "common/UsePrefixedDocumentTitle";
import {
GenerateTtsAudio,
GenerateTtsAudioRequest,
GenerateTtsAudioResponse,
GenerateTtsAudioIsOk,
} from "@storyteller/components/src/api/tts/GenerateTtsAudio";
import { v4 as uuidv4 } from "uuid";
import { WeightType } from "@storyteller/components/src/api/_common/enums";
import useWeightTypeInfo from "hooks/useWeightTypeInfo";
import { FrontendInferenceJobType } from "@storyteller/components/src/jobs/InferenceJob";
interface Props {
value?: any;
sessionSubscriptionsWrapper: any;
}
export default function DevTTS({ value }: Props) {
export default function DevTTS({ sessionSubscriptionsWrapper }: Props) {
const { enqueueInferenceJob } = useInferenceJobs();
const { modalState, open } = useModal();
const [search,searchSet] = useState("");
const [updated,updatedSet] = useState(false);
const [selectedVoice,selectedVoiceSet] = useState<Weight | undefined>();
const [text,textSet] = useState("");
const [search, searchSet] = useState("");
const [updated, updatedSet] = useState(false);
const [selectedVoice, selectedVoiceSet] = useState<Weight | undefined>();
const [text, textSet] = useState("");
const bucketConfig = new BucketConfig();
const preview = bucketConfig.getCdnUrl(selectedVoice?.cover_image?.maybe_cover_image_public_bucket_path || "");
const preview = selectedVoice?.cover_image
?.maybe_cover_image_public_bucket_path
? bucketConfig.getCdnUrl(
selectedVoice?.cover_image?.maybe_cover_image_public_bucket_path
)
: "/images/avatars/default-pfp.png";
const textChange = ({ target }: { target: any }) => {
textSet(target.value);
};
const [isGenerating, setIsGenerating] = useState(false);
usePrefixedDocumentTitle(
selectedVoice
? selectedVoice.title + " Text to Speech"
: "FakeYou. Deep Fake Text to Speech."
);
const searchChange = (setUpdate = true) => ({ target }: { target: any }) => {
// console.log("🩵",target.value);
if (setUpdate) updatedSet(true);
// console.log("aaaa",);
searchSet(target.value);
};
const weightTypeInfo = useWeightTypeInfo(
selectedVoice?.weight_type || WeightType.NONE
);
const { label: weightType, color: weightTagColor } = weightTypeInfo;
const searchChange =
(setUpdate = true) =>
({ target }: { target: any }) => {
if (setUpdate) updatedSet(true);
searchSet(target.value);
};
useDebounce({
blocked: !(updated && !modalState && search),
@@ -45,47 +90,257 @@ export default function DevTTS({ value }: Props) {
onSelect: (weight: any) => selectedVoiceSet(weight),
inputMode: 3,
onSearchChange: searchChange(false),
search
}
search,
},
});
}
},
});
return <Container>
<PageHeader {...{ title: "TTS" }}/>
<Panel padding {...{ className: ""}}>
<Label {...{ label: "Select a voice" }}/>
<div {...{ className: `fy-weight-picker mb-3 fy-weight-picker-${ selectedVoice ? "full" : "empty" }` }}>
<Input autoFocus {...{ icon: faSearch, placeholder: "Search 9999 voices...", onChange: searchChange(), value: search }}/>
{
selectedVoice && <div {...{ className: "fy-weight-picker-preview p-3" }}>
<WeightCoverImage {...{ src: preview, height: 60, width: 60 }}/>
<div {...{ className: "fy-weight-picker-preview-details" }}>
<h6>
{ selectedVoice.title }
<CardBadge {...{
className: `fy-entity-type-${ selectedVoice.weight_type || "" }`,
label: selectedVoice.weight_type || ""
}}/>
</h6>
<span>
by
<Link {...{ className: "fw-medium", to: `/profile/${ selectedVoice.creator?.username || "" }`}}>
{
" " + selectedVoice.creator?.display_name || ""
}
</Link>
</span>
const openModal = () => {
open({
component: MediaBrowser,
props: {
onSelect: (weight: any) => selectedVoiceSet(weight),
inputMode: 3,
onSearchChange: searchChange(false),
search,
accept: ["tts"],
title: "Select a Voice",
},
});
};
</div>
<Icon {...{ className: "fy-weight-picker-reselect", icon: faRotateLeft }}/>
</div>
}
</div>
{ selectedVoice && <TextArea autoFocus {...{ label: "Enter text", onChange: textChange, value: text }}/> }
{
text && <Button autoFocus {...{ label: "Speak" }}/>
const featuredVoiceTokens = [
"weight_31ewdsvev9bttgb4eg7zy7mj5",
"weight_b8rncypy7gw6nb0wthnwe2kk4",
"weight_3k28fws0v6r1ke3p0w0vw48gm",
"weight_0f762jdzgsy1dhpb86qxy4ssm",
"weight_1ptwk6pa8krh3ykfr7rztf3pz",
"weight_b8rncypy7gw6nb0wthnwe2kk4",
"weight_3k28fws0v6r1ke3p0w0vw48gm",
"weight_0f762jdzgsy1dhpb86qxy4ssm",
];
const handleSpeak = async () => {
if (!selectedVoice || !text) return;
setIsGenerating(true);
const request: GenerateTtsAudioRequest = {
uuid_idempotency_token: uuidv4(),
tts_model_token: selectedVoice.weight_token,
inference_text: text,
};
try {
const response: GenerateTtsAudioResponse =
await GenerateTtsAudio(request);
if (GenerateTtsAudioIsOk(response)) {
console.log("TTS queued successfully:", response.inference_job_token);
enqueueInferenceJob(
response.inference_job_token,
FrontendInferenceJobType.TextToSpeech
);
setIsGenerating(false);
} else {
console.error("Error queuing TTS:", response.error);
setIsGenerating(false);
}
</Panel>
</Container>;
};
} catch (error) {
console.error("Unexpected error:", error);
setIsGenerating(false);
}
};
return (
<>
<Container type="panel" className="mt-3 mt-lg-5">
<Panel padding={true}>
<h1 className="fw-bold mb-1">Text to Speech</h1>
<p className="mb-4 opacity-75 fw-medium">
Make your favorite characters speak!
</p>
<div className="d-flex flex-column gap-4">
<div className="fy-featured-voices-section d-none d-lg-block">
<h5 className="fw-bold">Featured Voices</h5>
<div className="row g-3">
{featuredVoiceTokens.map(token => (
<FeaturedVoice
key={token}
token={token}
onClick={selectedVoiceSet}
/>
))}
</div>
</div>
<div className="row">
<div className="d-flex flex-column gap-3 col-12 col-lg-6">
<div>
<div className="d-flex align-items-center">
{!selectedVoice && (
<div className="mb-2">
<div className="focus-point" />
</div>
)}
<div className="d-flex gap-2 align-items-center w-100">
<div className="flex-grow-1">
<Label
label={`${
selectedVoice ? "Selected Voice" : "Select a Voice"
}`}
/>
</div>
<div className="d-flex gap-2">
{selectedVoice && (
<Button
icon={faBell}
variant="link"
label="Notify me when this voice improves"
className="fs-7"
/>
)}
</div>
</div>
</div>
<div className="position-relative">
<Input
autoFocus
icon={faSearch}
placeholder={"Search from 3000+ voices"}
onChange={searchChange()}
value={search}
style={{ borderRadius: "0.5rem 0.5rem 0 0" }}
/>
{search && (
<FontAwesomeIcon
icon={faXmark}
className="position-absolute opacity-75 fs-5"
style={{
right: "1rem",
top: "50%",
transform: "translateY(-50%)",
cursor: "pointer",
}}
onClick={() => searchSet("")}
/>
)}
</div>
<div className="fy-weight-picker-preview" onClick={openModal}>
<WeightCoverImage
{...{ src: preview, height: 80, width: 80 }}
/>
<div className="d-flex flex-column justify-content-center flex-grow-1">
<h6 className="mb-1 fw-semibold d-flex gap-2 align-items-center">
<div>{selectedVoice?.title || "No Voice Selected"}</div>
{selectedVoice?.weight_type && (
<CardBadge
className={`fy-entity-type-${
selectedVoice?.weight_type || ""
}`}
label={weightType || ""}
small={true}
color={weightTagColor || ""}
/>
)}
</h6>
{selectedVoice ? (
<span className="fs-7 d-flex gap-1 flex-column flex-lg-row">
<div className="d-flex gap-1">
by
<Link
className="fw-medium"
to={`/profile/${
selectedVoice?.creator?.username || ""
}`}
onClick={e => e.stopPropagation()}
>
{" " + selectedVoice?.creator?.display_name || ""}
</Link>
</div>
{/* <span className="opacity-25">|</span>
<span>English</span> */}
<div className="d-flex gap-1 align-items-center">
<span className="d-none d-lg-block px-1 opacity-50">
|
</span>
<Link
to={`/weight/${selectedVoice.weight_token}`}
className="fw-medium"
onClick={e => e.stopPropagation()}
>
View voice details
<FontAwesomeIcon
icon={faArrowRight}
className="ms-2"
/>
</Link>
</div>
</span>
) : (
<span className="fs-7 opacity-75">
Click to select a voice
</span>
)}
</div>
<FontAwesomeIcon
icon={faChevronRight}
className="fs-5 me-1"
/>
</div>
</div>
<TextArea
label="Enter Text"
onChange={textChange}
value={text}
rows={5}
placeholder={`Enter the text you want ${
selectedVoice ? selectedVoice.title : "your character"
} to say...`}
resize={false}
/>
<div className="d-flex justify-content-end gap-2">
<Button
icon={faDeleteLeft}
label="Clear"
disabled={!text}
variant="secondary"
onClick={() => textSet("")}
/>
<Button
icon={faWaveformLines}
label="Speak"
onClick={handleSpeak}
disabled={!selectedVoice || !text}
isLoading={isGenerating}
/>
</div>
</div>
<div className="col-12 col-lg-6">
<div className="d-flex flex-column">
<Label label="Output" />
<div className="d-flex flex-column session-tts-section">
<SessionTtsInferenceResultList
sessionSubscriptionsWrapper={sessionSubscriptionsWrapper}
/>
</div>
</div>
</div>
</div>
</div>
</Panel>
</Container>
<Container type="panel" className="py-5 mt-5 d-flex flex-column gap-5">
{/* <MentionsSection /> */}
<StorytellerStudioCTA />.
</Container>
</>
);
}
@@ -0,0 +1,81 @@
import { WeightType } from "@storyteller/components/src/api/_common/enums";
import { BucketConfig } from "@storyteller/components/src/api/BucketConfig";
import LoadingSpinner from "components/common/LoadingSpinner";
import WeightCoverImage from "components/common/WeightCoverImage";
import { CardBadge } from "components/entities";
import { useWeightFetch } from "hooks";
import useWeightTypeInfo from "hooks/useWeightTypeInfo";
import React, { useState } from "react";
import { Link } from "react-router-dom";
interface FeaturedVoiceProps {
token: string;
onClick?: (weight: any) => void;
}
export const FeaturedVoice = ({ token, onClick }: FeaturedVoiceProps) => {
const weight_token = token;
const fetchedWeight = useWeightFetch({
onSuccess: () => setIsLoading(false),
token: weight_token,
});
const { data: weight } = fetchedWeight;
const bucketConfig = new BucketConfig();
const weightTypeInfo = useWeightTypeInfo(
weight?.weight_type || WeightType.NONE
);
const { label: weightType, color: weightTagColor } = weightTypeInfo;
const coverImage = weight?.cover_image?.maybe_cover_image_public_bucket_path
? bucketConfig.getCdnUrl(
weight?.cover_image?.maybe_cover_image_public_bucket_path
)
: "/images/avatars/default-pfp.png";
const [isLoading, setIsLoading] = useState(true);
return (
<div className="col-12 col-lg-6 col-xl-3">
<div
className="fy-featured-voices d-flex align-items-center position-relative"
onClick={() => onClick && weight && onClick(weight)}
style={{ height: "74px" }}
>
{isLoading ? (
<div className="w-100 d-flex align-items-center justify-content-center">
<LoadingSpinner />
</div>
) : (
<>
<WeightCoverImage {...{ src: coverImage, height: 50, width: 50 }} />
<div className="d-flex flex-column justify-content-center">
<h6 className="mb-1 fw-semibold d-flex align-items-center gap-2">
{weight?.title}
<CardBadge
label={weightType}
color={weightTagColor}
small={true}
/>
</h6>
<span className="fs-7 fw-medium">
by{" "}
<Link to={`/profile/${weight?.creator.display_name}`}>
{weight?.creator.display_name}
</Link>
</span>
</div>
<div
className="position-absolute fs-7 fw-medium fy-select-voice"
style={{ bottom: "8px", right: "8px" }}
>
Use
</div>
</>
)}
</div>
</div>
);
};