summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authoryum <yum.food.vr@gmail.com>2023-02-24 14:35:02 -0800
committeryum <yum.food.vr@gmail.com>2023-02-24 14:35:02 -0800
commit8170273eb786d52aa9a9c1cdbef1357c32d95471 (patch)
tree0a19c8632beec6e51e51582a004bed21b0e04041
parentc18b6842757e1a40f27bd940a810e32f3a0c485c (diff)
Finish browser source proof-of-concept
It's a crashy mess, but it sort of works. * Add Transcript class to send transcription segments between layers
Diffstat
-rw-r--r--BrowserSource/index.html13
-rw-r--r--GUI/GUI/GUI/BrowserSource.cpp6
-rw-r--r--GUI/GUI/GUI/BrowserSource.h21
-rw-r--r--GUI/GUI/GUI/GUI.vcxproj2
-rw-r--r--GUI/GUI/GUI/GUI.vcxproj.filters6
-rw-r--r--GUI/GUI/GUI/Transcript.cpp16
-rw-r--r--GUI/GUI/GUI/Transcript.h20
-rw-r--r--GUI/GUI/GUI/WhisperCPP.cpp16
-rw-r--r--GUI/GUI/GUI/WhisperCPP.h3
9 files changed, 88 insertions, 15 deletions
diff --git a/BrowserSource/index.html b/BrowserSource/index.html
index decdba2..29cffbf 100644
--- a/BrowserSource/index.html
+++ b/BrowserSource/index.html
@@ -3,10 +3,22 @@
<head>
<title>TaSTT</title>
<script src="https://code.jquery.com/jquery-3.6.3.min.js"></script>
+ <link rel="stylesheet"
+ href="https://fonts.googleapis.com/css?family=Noto+Sans+Mono&display=swap">
</head>
+ <style>
+ body {
+ font-family: 'Noto Sans Mono', monospace;
+ font-size: 48px;
+ font-weight: 700;
+ }
+ </style>
<body>
<div id="transcript"></div>
<script>
+ function scrollToBottom() {
+ window.scrollTo(0,document.body.scrollHeight);
+ }
function getTranscript() {
$.ajax({
// TODO(yum) parameterize the port
@@ -21,6 +33,7 @@
console.error('Error getting transcript: ', textStatus, errorThrown);
}
});
+ scrollToBottom();
}
setInterval(getTranscript, /*interval_ms=*/100);
</script>
diff --git a/GUI/GUI/GUI/BrowserSource.cpp b/GUI/GUI/GUI/BrowserSource.cpp
index 9bf19cb..2403411 100644
--- a/GUI/GUI/GUI/BrowserSource.cpp
+++ b/GUI/GUI/GUI/BrowserSource.cpp
@@ -6,8 +6,8 @@
using ::Logging::Log;
-BrowserSource::BrowserSource(uint16_t port, wxTextCtrl *out)
- : port_(port), out_(out)
+BrowserSource::BrowserSource(uint16_t port, wxTextCtrl *out, Transcript *transcript)
+ : port_(port), out_(out), transcript_(transcript)
{}
void BrowserSource::Run(volatile bool* run)
@@ -37,7 +37,7 @@ void BrowserSource::Run(volatile bool* run)
OATPP_CREATE_COMPONENT(std::shared_ptr<oatpp::web::server::HttpRouter>, httpRouter)([] {
return oatpp::web::server::HttpRouter::createShared();
}());
- httpRouter.getObject()->addController(std::make_shared<AppController>(apiObjectMapper.getObject()));
+ httpRouter.getObject()->addController(std::make_shared<AppController>(apiObjectMapper.getObject(), transcript_));
OATPP_CREATE_COMPONENT(std::shared_ptr<oatpp::network::ConnectionHandler>, serverConnectionHandler)([&] {
return oatpp::web::server::HttpConnectionHandler::createShared(httpRouter.getObject());
diff --git a/GUI/GUI/GUI/BrowserSource.h b/GUI/GUI/GUI/BrowserSource.h
index 6387d1c..25b9aa5 100644
--- a/GUI/GUI/GUI/BrowserSource.h
+++ b/GUI/GUI/GUI/BrowserSource.h
@@ -15,6 +15,8 @@
#include "oatpp/web/server/HttpConnectionHandler.hpp"
#include "oatpp/web/protocol/http/incoming/Request.hpp"
+#include "Transcript.h"
+
#include <stdint.h>
#include <filesystem>
@@ -37,15 +39,20 @@ class AppDto : public oatpp::DTO
class AppController : public oatpp::web::server::api::ApiController
{
public:
- AppController(std::shared_ptr<ObjectMapper> objectMapper)
- : oatpp::web::server::api::ApiController(objectMapper)
+ AppController(std::shared_ptr<ObjectMapper> objectMapper, Transcript* transcript)
+ : oatpp::web::server::api::ApiController(objectMapper), transcript_(transcript)
{}
-public:
ENDPOINT("GET", "/api/transcript", transcription) {
auto dto = AppDto::createShared();
dto->statusCode = 200;
- dto->transcript = "Hello World!";
+
+ std::ostringstream oss;
+ std::vector<std::string> segments = transcript_->Get();
+ for (const auto& seg : segments) {
+ oss << seg;
+ }
+ dto->transcript = oss.str();
return createDtoResponse(Status::CODE_200, dto);
}
@@ -57,6 +64,9 @@ public:
html_ifs.read(resp.data(), resp.size());
return createResponse(Status::CODE_200, resp.data());
}
+
+private:
+ Transcript* const transcript_;
};
#include OATPP_CODEGEN_END(ApiController)
@@ -64,12 +74,13 @@ public:
class BrowserSource
{
public:
- BrowserSource(uint16_t port, wxTextCtrl *out);
+ BrowserSource(uint16_t port, wxTextCtrl *out, Transcript *transcript);
void Run(volatile bool* run);
private:
const uint16_t port_;
wxTextCtrl* const out_;
+ Transcript* const transcript_;
};
diff --git a/GUI/GUI/GUI/GUI.vcxproj b/GUI/GUI/GUI/GUI.vcxproj
index e3fac34..568b06f 100644
--- a/GUI/GUI/GUI/GUI.vcxproj
+++ b/GUI/GUI/GUI/GUI.vcxproj
@@ -157,6 +157,7 @@
<ClCompile Include="Logging.cpp" />
<ClCompile Include="main.cpp" />
<ClCompile Include="PythonWrapper.cpp" />
+ <ClCompile Include="Transcript.cpp" />
<ClCompile Include="WhisperCPP.cpp" />
</ItemGroup>
<ItemGroup>
@@ -169,6 +170,7 @@
<ClInclude Include="resource.h" />
<ClInclude Include="ryml.h" />
<ClInclude Include="ScopeGuard.h" />
+ <ClInclude Include="Transcript.h" />
<ClInclude Include="Util.h" />
<ClInclude Include="WhisperCPP.h" />
</ItemGroup>
diff --git a/GUI/GUI/GUI/GUI.vcxproj.filters b/GUI/GUI/GUI/GUI.vcxproj.filters
index 2d87bd3..aa2e6d1 100644
--- a/GUI/GUI/GUI/GUI.vcxproj.filters
+++ b/GUI/GUI/GUI/GUI.vcxproj.filters
@@ -39,6 +39,9 @@
<ClCompile Include="BrowserSource.cpp">
<Filter>Source Files</Filter>
</ClCompile>
+ <ClCompile Include="Transcript.cpp">
+ <Filter>Source Files</Filter>
+ </ClCompile>
</ItemGroup>
<ItemGroup>
<ClInclude Include="PythonWrapper.h">
@@ -74,6 +77,9 @@
<ClInclude Include="BrowserSource.h">
<Filter>Header Files</Filter>
</ClInclude>
+ <ClInclude Include="Transcript.h">
+ <Filter>Header Files</Filter>
+ </ClInclude>
</ItemGroup>
<ItemGroup>
<ResourceCompile Include="GUI.rc">
diff --git a/GUI/GUI/GUI/Transcript.cpp b/GUI/GUI/GUI/Transcript.cpp
new file mode 100644
index 0000000..30f1f76
--- /dev/null
+++ b/GUI/GUI/GUI/Transcript.cpp
@@ -0,0 +1,16 @@
+#include "Transcript.h"
+
+void Transcript::Append(std::string&& segment) {
+ std::scoped_lock l(mu_);
+ segments_.push_back(std::move(segment));
+}
+
+void Transcript::Clear() {
+ std::scoped_lock l(mu_);
+ segments_.clear();
+}
+
+std::vector<std::string> Transcript::Get() {
+ std::scoped_lock l(mu_);
+ return segments_;
+}
diff --git a/GUI/GUI/GUI/Transcript.h b/GUI/GUI/GUI/Transcript.h
new file mode 100644
index 0000000..09858b0
--- /dev/null
+++ b/GUI/GUI/GUI/Transcript.h
@@ -0,0 +1,20 @@
+#pragma once
+
+#include <mutex>
+#include <string>
+#include <vector>
+
+// Simple thread-safe class to share transcription data between layers.
+class Transcript {
+public:
+ Transcript() = default;
+
+ void Append(std::string&& segment);
+ void Clear();
+
+ std::vector<std::string> Get();
+
+private:
+ std::mutex mu_;
+ std::vector<std::string> segments_;
+};
diff --git a/GUI/GUI/GUI/WhisperCPP.cpp b/GUI/GUI/GUI/WhisperCPP.cpp
index 2e872f9..e8ed4ef 100644
--- a/GUI/GUI/GUI/WhisperCPP.cpp
+++ b/GUI/GUI/GUI/WhisperCPP.cpp
@@ -238,6 +238,7 @@ bool WhisperCPP::CreateContext(Whisper::iModel* model, Whisper::iContext*& conte
void WhisperCPP::Start(const AppConfig& c) {
Init();
+ transcript_.Clear();
if (!transcription_thd_.valid()) {
Log(out_, "Transcription engine already running\n");
@@ -301,11 +302,11 @@ void WhisperCPP::Start(const AppConfig& c) {
wparams.n_max_text_ctx = 100;
wparams.new_segment_callback = [](iContext* context, uint32_t n_new, void* user_data) noexcept -> HRESULT {
- wxTextCtrl* out = static_cast<wxTextCtrl*>(user_data);
+ WhisperCPP* app = static_cast<WhisperCPP*>(user_data);
iTranscribeResult* results = nullptr;
HRESULT err = context->getResults(eResultFlags::Timestamps | eResultFlags::Tokens, &results);
if (FAILED(err)) {
- Log(out, "Failed to get transcription: {}\n", hresultToString(err));
+ Log(app->out_, "Failed to get transcription: {}\n", hresultToString(err));
return S_OK;
}
ScopeGuard results_cleanup([results]() { results->Release(); });
@@ -313,7 +314,7 @@ void WhisperCPP::Start(const AppConfig& c) {
sTranscribeLength length;
err = results->getSize(length);
if (FAILED(err)) {
- Log(out, "Failed to get transcription size: {}\n", hresultToString(err));
+ Log(app->out_, "Failed to get transcription size: {}\n", hresultToString(err));
return S_OK;
}
@@ -366,16 +367,17 @@ void WhisperCPP::Start(const AppConfig& c) {
continue;
}
#endif
- Log(out, "{}", tok.text);
+ Log(app->out_, "{}", tok.text);
+ app->transcript_.Append(tok.text);
}
}
if (n_new) {
- Log(out, "\n");
+ Log(app->out_, "\n");
}
return S_OK;
};
- wparams.new_segment_callback_user_data = out_;
+ wparams.new_segment_callback_user_data = this;
sCaptureCallbacks callbacks{};
@@ -418,7 +420,7 @@ void WhisperCPP::StartBrowserSource(const AppConfig& c) {
browser_src_thd_ = std::async(std::launch::async, [&]() -> void {
run_browser_src_ = true;
- BrowserSource src(c.browser_src_port, out_);
+ BrowserSource src(c.browser_src_port, out_, &transcript_);
src.Run(&run_browser_src_);
Log(out_, "Browser source thread exit\n");
});
diff --git a/GUI/GUI/GUI/WhisperCPP.h b/GUI/GUI/GUI/WhisperCPP.h
index c6e1011..fbaab46 100644
--- a/GUI/GUI/GUI/WhisperCPP.h
+++ b/GUI/GUI/GUI/WhisperCPP.h
@@ -13,6 +13,7 @@
#include "whisper/whisperWindows.h"
#include "Config.h"
+#include "Transcript.h"
#include <filesystem>
#include <functional>
@@ -52,4 +53,6 @@ private:
std::future<void> browser_src_thd_;
volatile bool run_browser_src_;
+
+ Transcript transcript_;
};
generated by cgit v1.2.3 (git 2.39.1) at 2026-06-01 14:34:05 +0000