From cad0f796fa4dab21691fa7f0971b5c180bdcf242 Mon Sep 17 00:00:00 2001 From: Sakamoto Shiina <68018796+ShiinaSakamoto@users.noreply.github.com> Date: Fri, 5 Sep 2025 10:44:23 +0900 Subject: [PATCH] [Update] Transcription: Add UI. The user can config 'mic/speaker avg logprob' and 'mic/speaker no speech prob' that is related with Whisper's parameter. --- .../setting_box/appearance/Appearance.jsx | 3 + .../transcription/Transcription.jsx | 195 ++++++++++++++++++ .../configs/transcription/useTranscription.js | 99 +++++++++ src-ui/logics/useReceiveRoutes.js | 10 + src-ui/store.js | 5 + 5 files changed, 312 insertions(+) diff --git a/src-ui/app/config_page/setting_section/setting_box/appearance/Appearance.jsx b/src-ui/app/config_page/setting_section/setting_box/appearance/Appearance.jsx index 1f2f9eb4..cc7e730f 100644 --- a/src-ui/app/config_page/setting_section/setting_box/appearance/Appearance.jsx +++ b/src-ui/app/config_page/setting_section/setting_box/appearance/Appearance.jsx @@ -69,6 +69,7 @@ const UiScalingContainer = () => { asyncUpdateBreakPoint(); }, [currentUiScaling.data]); + // [Duplicated] const createMarks = (min, max) => { const marks = []; for (let value = min; value <= max; value += 10) { @@ -111,6 +112,7 @@ export const MessageLogUiScalingContainer = () => { setUiMessageLogUiScaling(currentMessageLogUiScaling.data); }, [currentMessageLogUiScaling.data]); + // [Duplicated] const createMarks = (min, max) => { const marks = []; for (let value = min; value <= max; value += 10) { @@ -207,6 +209,7 @@ const TransparencyContainer = () => { setUiTransparency(currentTransparency.data); }, [currentTransparency.data]); + // [Duplicated] const createMarks = (min, max) => { const marks = []; for (let value = min; value <= max; value += 10) { diff --git a/src-ui/app/config_page/setting_section/setting_box/transcription/Transcription.jsx b/src-ui/app/config_page/setting_section/setting_box/transcription/Transcription.jsx index 279c1d9d..7416c970 100644 --- a/src-ui/app/config_page/setting_section/setting_box/transcription/Transcription.jsx +++ b/src-ui/app/config_page/setting_section/setting_box/transcription/Transcription.jsx @@ -1,3 +1,4 @@ +import { useEffect, useState } from "react"; import { useI18n } from "@useI18n"; import styles from "./Transcription.module.scss"; import { updateLabelsById, genNumObjArray } from "@utils"; @@ -12,6 +13,7 @@ import { RadioButtonContainer, DropdownMenuContainer, ComputeDeviceContainer, + SliderContainer, } from "../_templates/Templates"; import { @@ -24,6 +26,7 @@ export const Transcription = () => { + ); }; @@ -353,4 +356,196 @@ const findKeyByDeviceValue = (devices, target_value) => { } } return null; +}; + + + + + +const Advanced_Container = () => { + const { t } = useI18n(); + return ( +
+ + {/* */} + + + + +
+ ); + + +}; + +export const MicAvgLogprobContainer = () => { + const { t } = useI18n(); + const { currentMicAvgLogprob, setMicAvgLogprob } = useTranscription(); + const [ui_mic_avg_logprob, setUiMicAvgLogprob] = useState(currentMicAvgLogprob.data); + + const onchangeFunction = (value) => { + setUiMicAvgLogprob(value); + }; + const onchangeCommittedFunction = (value) => { + setMicAvgLogprob(value); + }; + useEffect(() => { + setUiMicAvgLogprob(currentMicAvgLogprob.data); + }, [currentMicAvgLogprob.data]); + + // [Duplicated] + const createMarks = (min, max) => { + const marks = []; + for (let value = min; value <= max; value += 0.2) { + value = parseFloat(value.toFixed(1)); + marks.push({ value, label: `${value}` }); + } + return marks; + }; + + const marks = createMarks(-2, 0); + + return ( + + ); +}; + +export const MicNoSpeechProbContainer = () => { + const { t } = useI18n(); + const { currentMicNoSpeechProb, setMicNoSpeechProb } = useTranscription(); + const [ui_mic_no_speech_prob, setUiMicNoSpeechProb] = useState(currentMicNoSpeechProb.data); + + const onchangeFunction = (value) => { + setUiMicNoSpeechProb(value); + }; + const onchangeCommittedFunction = (value) => { + setMicNoSpeechProb(value); + }; + useEffect(() => { + setUiMicNoSpeechProb(currentMicNoSpeechProb.data); + }, [currentMicNoSpeechProb.data]); + + // [Duplicated] + const createMarks = (min, max) => { + const marks = []; + for (let value = min; value <= max; value += 0.1) { + value = parseFloat(value.toFixed(1)); + marks.push({ value, label: `${value}` }); + } + return marks; + }; + + const marks = createMarks(0, 1); + + return ( + + ); +}; + +export const SpeakerAvgLogprobContainer = () => { + const { t } = useI18n(); + const { currentSpeakerAvgLogprob, setSpeakerAvgLogprob } = useTranscription(); + const [ui_speaker_avg_logprob, setUiSpeakerAvgLogprob] = useState(currentSpeakerAvgLogprob.data); + + const onchangeFunction = (value) => { + setUiSpeakerAvgLogprob(value); + }; + const onchangeCommittedFunction = (value) => { + setSpeakerAvgLogprob(value); + }; + useEffect(() => { + setUiSpeakerAvgLogprob(currentSpeakerAvgLogprob.data); + }, [currentSpeakerAvgLogprob.data]); + + // [Duplicated] + const createMarks = (min, max) => { + const marks = []; + for (let value = min; value <= max; value += 0.2) { + value = parseFloat(value.toFixed(1)); + marks.push({ value, label: `${value}` }); + } + return marks; + }; + + const marks = createMarks(-2, 0); + + return ( + + ); +}; + +export const SpeakerNoSpeechProbContainer = () => { + const { t } = useI18n(); + const { currentSpeakerNoSpeechProb, setSpeakerNoSpeechProb } = useTranscription(); + const [ui_speaker_no_speech_prob, setUiSpeakerNoSpeechProb] = useState(currentSpeakerNoSpeechProb.data); + + const onchangeFunction = (value) => { + setUiSpeakerNoSpeechProb(value); + }; + const onchangeCommittedFunction = (value) => { + setSpeakerNoSpeechProb(value); + }; + useEffect(() => { + setUiSpeakerNoSpeechProb(currentSpeakerNoSpeechProb.data); + }, [currentSpeakerNoSpeechProb.data]); + + // [Duplicated] + const createMarks = (min, max) => { + const marks = []; + for (let value = min; value <= max; value += 0.1) { + value = parseFloat(value.toFixed(1)); + marks.push({ value, label: `${value}` }); + } + return marks; + }; + + const marks = createMarks(0, 1); + + return ( + + ); }; \ No newline at end of file diff --git a/src-ui/logics/configs/transcription/useTranscription.js b/src-ui/logics/configs/transcription/useTranscription.js index 147aa9e6..294b0473 100644 --- a/src-ui/logics/configs/transcription/useTranscription.js +++ b/src-ui/logics/configs/transcription/useTranscription.js @@ -14,6 +14,11 @@ import { useStore_SelectedWhisperWeightType, useStore_WhisperWeightTypeStatus, + + useStore_MicAvgLogprob, + useStore_MicNoSpeechProb, + useStore_SpeakerAvgLogprob, + useStore_SpeakerNoSpeechProb, } from "@store"; import { useStdoutToPython } from "@useStdoutToPython"; import { transformToIndexedArray } from "@utils"; @@ -41,6 +46,13 @@ export const useTranscription = () => { const { currentSelectableWhisperComputeDeviceList, updateSelectableWhisperComputeDeviceList, pendingSelectableWhisperComputeDeviceList } = useStore_SelectableWhisperComputeDeviceList(); const { currentSelectedWhisperComputeDevice, updateSelectedWhisperComputeDevice, pendingSelectedWhisperComputeDevice } = useStore_SelectedWhisperComputeDevice(); + // Advanced Settings + const { currentMicAvgLogprob, updateMicAvgLogprob, pendingMicAvgLogprob } = useStore_MicAvgLogprob(); + const { currentMicNoSpeechProb, updateMicNoSpeechProb, pendingMicNoSpeechProb } = useStore_MicNoSpeechProb(); + const { currentSpeakerAvgLogprob, updateSpeakerAvgLogprob, pendingSpeakerAvgLogprob } = useStore_SpeakerAvgLogprob(); + const { currentSpeakerNoSpeechProb, updateSpeakerNoSpeechProb, pendingSpeakerNoSpeechProb } = useStore_SpeakerNoSpeechProb(); + + // Mic const getMicRecordTimeout = () => { pendingMicRecordTimeout(); @@ -276,6 +288,67 @@ export const useTranscription = () => { showNotification_SaveSuccess(); }; + // Advanced (Mic Avg Logprob) + const getMicAvgLogprob = () => { + pendingMicAvgLogprob(); + asyncStdoutToPython("/get/data/mic_avg_logprob"); + }; + + const setMicAvgLogprob = (selected_mic_avg_logprob) => { + pendingMicAvgLogprob(); + asyncStdoutToPython("/set/data/mic_avg_logprob", selected_mic_avg_logprob); + }; + + const setSuccessMicAvgLogprob = (selected_mic_avg_logprob) => { + updateMicAvgLogprob(selected_mic_avg_logprob); + showNotification_SaveSuccess(); + }; + // Advanced (Mic No Speech Prob) + const getMicNoSpeechProb = () => { + pendingMicNoSpeechProb(); + asyncStdoutToPython("/get/data/mic_no_speech_prob"); + }; + + const setMicNoSpeechProb = (selected_mic_no_speech_prob) => { + pendingMicNoSpeechProb(); + asyncStdoutToPython("/set/data/mic_no_speech_prob", selected_mic_no_speech_prob); + }; + + const setSuccessMicNoSpeechProb = (selected_mic_no_speech_prob) => { + updateMicNoSpeechProb(selected_mic_no_speech_prob); + showNotification_SaveSuccess(); + }; + // Advanced (Speaker Avg Logprob) + const getSpeakerAvgLogprob = () => { + pendingSpeakerAvgLogprob(); + asyncStdoutToPython("/get/data/speaker_avg_logprob"); + }; + + const setSpeakerAvgLogprob = (selected_speaker_avg_logprob) => { + pendingSpeakerAvgLogprob(); + asyncStdoutToPython("/set/data/speaker_avg_logprob", selected_speaker_avg_logprob); + }; + + const setSuccessSpeakerAvgLogprob = (selected_speaker_avg_logprob) => { + updateSpeakerAvgLogprob(selected_speaker_avg_logprob); + showNotification_SaveSuccess(); + }; + // Advanced (Speaker No Speech Prob) + const getSpeakerNoSpeechProb = () => { + pendingSpeakerNoSpeechProb(); + asyncStdoutToPython("/get/data/speaker_no_speech_prob"); + }; + + const setSpeakerNoSpeechProb = (selected_speaker_no_speech_prob) => { + pendingSpeakerNoSpeechProb(); + asyncStdoutToPython("/set/data/speaker_no_speech_prob", selected_speaker_no_speech_prob); + }; + + const setSuccessSpeakerNoSpeechProb = (selected_speaker_no_speech_prob) => { + updateSpeakerNoSpeechProb(selected_speaker_no_speech_prob); + showNotification_SaveSuccess(); + }; + return { // Mic currentMicRecordTimeout, @@ -353,5 +426,31 @@ export const useTranscription = () => { updateSelectedWhisperComputeDevice, setSelectedWhisperComputeDevice, setSuccessSelectedWhisperComputeDevice, + + // Advanced + // Mic Avg Logprob + currentMicAvgLogprob, + getMicAvgLogprob, + updateMicAvgLogprob, + setMicAvgLogprob, + setSuccessMicAvgLogprob, + // Mic No Speech Prob + currentMicNoSpeechProb, + getMicNoSpeechProb, + updateMicNoSpeechProb, + setMicNoSpeechProb, + setSuccessMicNoSpeechProb, + // Speaker Avg Logprob + currentSpeakerAvgLogprob, + getSpeakerAvgLogprob, + updateSpeakerAvgLogprob, + setSpeakerAvgLogprob, + setSuccessSpeakerAvgLogprob, + // Speaker No Speech Prob + currentSpeakerNoSpeechProb, + getSpeakerNoSpeechProb, + updateSpeakerNoSpeechProb, + setSpeakerNoSpeechProb, + setSuccessSpeakerNoSpeechProb, }; }; \ No newline at end of file diff --git a/src-ui/logics/useReceiveRoutes.js b/src-ui/logics/useReceiveRoutes.js index c9380612..c36b5fa7 100644 --- a/src-ui/logics/useReceiveRoutes.js +++ b/src-ui/logics/useReceiveRoutes.js @@ -224,6 +224,16 @@ export const ROUTE_META_LIST = [ { endpoint: "/get/data/selected_transcription_compute_device", ns: configs, hook_name: "useTranscription", method_name: "updateSelectedWhisperComputeDevice" }, { endpoint: "/set/data/selected_transcription_compute_device", ns: configs, hook_name: "useTranscription", method_name: "setSuccessSelectedWhisperComputeDevice" }, + // Transcription (Advanced) + { endpoint: "/get/data/mic_avg_logprob", ns: configs, hook_name: "useTranscription", method_name: "updateMicAvgLogprob" }, + { endpoint: "/set/data/mic_avg_logprob", ns: configs, hook_name: "useTranscription", method_name: "setSuccessMicAvgLogprob" }, + { endpoint: "/get/data/mic_no_speech_prob", ns: configs, hook_name: "useTranscription", method_name: "updateMicNoSpeechProb" }, + { endpoint: "/set/data/mic_no_speech_prob", ns: configs, hook_name: "useTranscription", method_name: "setSuccessMicNoSpeechProb" }, + { endpoint: "/get/data/speaker_avg_logprob", ns: configs, hook_name: "useTranscription", method_name: "updateSpeakerAvgLogprob" }, + { endpoint: "/set/data/speaker_avg_logprob", ns: configs, hook_name: "useTranscription", method_name: "setSuccessSpeakerAvgLogprob" }, + { endpoint: "/get/data/speaker_no_speech_prob", ns: configs, hook_name: "useTranscription", method_name: "updateSpeakerNoSpeechProb" }, + { endpoint: "/set/data/speaker_no_speech_prob", ns: configs, hook_name: "useTranscription", method_name: "setSuccessSpeakerNoSpeechProb" }, + // VR { endpoint: "/get/data/overlay_small_log", ns: configs, hook_name: "useVr", method_name: "updateIsEnabledOverlaySmallLog" }, { endpoint: "/set/enable/overlay_small_log", ns: configs, hook_name: "useVr", method_name: "setSuccessIsEnabledOverlaySmallLog" }, diff --git a/src-ui/store.js b/src-ui/store.js index 9241ad41..e36aad46 100644 --- a/src-ui/store.js +++ b/src-ui/store.js @@ -238,6 +238,11 @@ export const { atomInstance: Atom_SelectedTranscriptionEngine, useHook: useStore export const { atomInstance: Atom_SelectableWhisperComputeDeviceList, useHook: useStore_SelectableWhisperComputeDeviceList } = createAtomWithHook({}, "SelectableWhisperComputeDeviceList"); export const { atomInstance: Atom_SelectedWhisperComputeDevice, useHook: useStore_SelectedWhisperComputeDevice } = createAtomWithHook("", "SelectedWhisperComputeDevice"); +export const { atomInstance: Atom_MicAvgLogprob, useHook: useStore_MicAvgLogprob } = createAtomWithHook(-0.8, "MicAvgLogprob"); +export const { atomInstance: Atom_MicNoSpeechProb, useHook: useStore_MicNoSpeechProb } = createAtomWithHook(0.6, "MicNoSpeechProb"); +export const { atomInstance: Atom_SpeakerAvgLogprob, useHook: useStore_SpeakerAvgLogprob } = createAtomWithHook(-0.8, "SpeakerAvgLogprob"); +export const { atomInstance: Atom_SpeakerNoSpeechProb, useHook: useStore_SpeakerNoSpeechProb } = createAtomWithHook(0.6, "SpeakerNoSpeechProb"); + // VR export const { atomInstance: Atom_OverlaySmallLogSettings, useHook: useStore_OverlaySmallLogSettings } = createAtomWithHook({