From: Georgi Gerganov <redacted>
Date: Wed, 23 Nov 2022 22:34:00 +0000 (+0200)
Subject: talk.wasm : add audio pre-processing + bump memory
X-Git-Tag: upstream/1.7.4~1794
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=37422ed7339eb8e2a3c827cea6d690ea16fb07e7;p=pkg%2Fggml%2Fsources%2Fwhisper.cpp

talk.wasm : add audio pre-processing + bump memory
---

diff --git a/examples/talk.wasm/CMakeLists.txt b/examples/talk.wasm/CMakeLists.txt
index 6731d62e..35f62234 100644
--- a/examples/talk.wasm/CMakeLists.txt
+++ b/examples/talk.wasm/CMakeLists.txt
@@ -31,8 +31,8 @@ set_target_properties(${TARGET} PROPERTIES LINK_FLAGS " \
     --bind \
     -s USE_PTHREADS=1 \
     -s PTHREAD_POOL_SIZE=8 \
-    -s INITIAL_MEMORY=1400MB \
-    -s TOTAL_MEMORY=1400MB \
+    -s INITIAL_MEMORY=1600MB \
+    -s TOTAL_MEMORY=1600MB \
     -s FORCE_FILESYSTEM=1 \
     -s EXPORTED_RUNTIME_METHODS=\"['print', 'printErr', 'ccall', 'cwrap']\" \
     ${EXTRA_FLAGS} \
diff --git a/examples/talk.wasm/README.md b/examples/talk.wasm/README.md
index c1b16c30..1ee272ad 100644
--- a/examples/talk.wasm/README.md
+++ b/examples/talk.wasm/README.md
@@ -34,7 +34,7 @@ In order to run this demo efficiently, you need to have the following:
 - Latest Chrome or Firefox browser (Safari is not supported)
 - Run this on a desktop or laptop with modern CPU (a mobile phone will likely not be good enough)
 - Speak phrases that are no longer than 10 seconds - this is the audio context of the AI
-- The web-page uses about 1.4GB of RAM
+- The web-page uses about 1.6GB of RAM
 
 Notice that this demo is using the smallest GPT-2 model, so the generated text responses are not always very good.
 Also, the prompting strategy can likely be improved to achieve better results.
diff --git a/examples/talk.wasm/gpt-2.cpp b/examples/talk.wasm/gpt-2.cpp
index 2e25c070..c6526e44 100644
--- a/examples/talk.wasm/gpt-2.cpp
+++ b/examples/talk.wasm/gpt-2.cpp
@@ -513,7 +513,7 @@ bool gpt2_eval(
     const int n_head  = hparams.n_head;
     const int n_vocab = hparams.n_vocab;
 
-    static size_t buf_size = 512u*1024*1024;
+    static size_t buf_size = 640u*1024*1024;
     static void * buf = malloc(buf_size);
 
     if (mem_per_token > 0 && mem_per_token*N > buf_size) {
diff --git a/examples/talk.wasm/index-tmpl.html b/examples/talk.wasm/index-tmpl.html
index c8c9e112..f07a57ca 100644
--- a/examples/talk.wasm/index-tmpl.html
+++ b/examples/talk.wasm/index-tmpl.html
@@ -504,7 +504,13 @@
 
             function startRecording() {
                 if (!context) {
-                    context = new AudioContext({sampleRate: 16000, noiseSuppression: true});
+                    context = new AudioContext({
+                        sampleRate: 16000,
+                        channelCount: 1,
+                        echoCancellation: true,
+                        autoGainControl:  true,
+                        noiseSuppression: true,
+                    });
                 }
 
                 Module.set_status("");