whisper large v3, arranque run.sh

whisper/lib/whisper_web/channels/audio_channel.ex

@@ -31,49 +31,69 @@ defmodule WhisperWeb.AudioChannel do
 
     Logger.info("Chunk recibido: #{byte_size(audio)} bytes, sample_rate: #{rate}")
     AudioBuffer.append(ref, {rate, audio})
+    chunks = AudioBuffer.get_and_clear(ref)
 
-    # {:ok, path} = AudioSaver.save_chunk_as_wav(ref, audio, rate, "part")
-    # AudioFilesList.add_file(path)
+    start_total = System.monotonic_time(:millisecond)
+
+    if chunks != [] do
+
+      Task.start(fn ->
+        [{rate, _} | _] = chunks
+        full_audio = Enum.map(chunks, fn {_, bin} -> bin end) |> IO.iodata_to_binary()
+
+        {wav_time, {:ok, path}} =
+          :timer.tc(fn ->
+            AudioSaver.save_chunk_as_wav(ref, full_audio, rate, "part")
+          end)
+        model_start = System.monotonic_time(:millisecond)
+        Logger.info("WAV guardado en #{div(wav_time, 1000)} ms")
 
 
+        transcription =
+          if path do
+            case Nx.Serving.batched_run(Whisper.LargeModel.Serving, {:file, path}) do
+              %{chunks: chunks} ->
+                chunks
+                |> Enum.map(& &1.text)
+                |> Enum.join(" ")
+
+              _ ->
+                "Transcripción no disponible"
+            end
+          else
+            "Archivo no disponible"
+          end
+
+
+        model_end = System.monotonic_time(:millisecond)
+        Logger.info("El modelo procesó en #{model_end - model_start} ms")
+
+        Logger.info("✅ Transcripción:\n#{transcription}")
+
+        message = %{"chunks" => [%{"text" => transcription}]}
+        PubSub.broadcast(Whisper.PubSub, "transcription", {:transcription, %{
+          "received_at" => model_start,
+          "text" => transcription
+        }})
+        File.rm!(path)
+
+        end_total = System.monotonic_time(:millisecond)
+        Logger.info("⏱ Total procesamiento stop_audio: #{end_total - start_total} ms")
+      end)
+    end
     {:noreply, socket}
   end
 
-
-  @doc """
-  Recupera todos los chunks acumulados en el buffer, los concatena y guarda un archivo WAV final (sufijo `"final"`).
-  """
   def handle_in("stop_audio", _payload, socket) do
-
     Logger.info("🛑 Grabación detenida por cliente")
 
     ref = socket_id(socket)
-    chunks = AudioBuffer.get_and_clear(ref)
 
-    if chunks != [] do
-      [{rate, _} | _] = chunks
-      full_audio = Enum.map(chunks, fn {_, bin} -> bin end) |> IO.iodata_to_binary()
-      {:ok, path} = AudioSaver.save_chunk_as_wav(ref, full_audio, rate, "final")
-      
-      Task.start(fn ->
-        transcription = Whisper.SendToModel.large(path)
-        Logger.info("✅ Transcripción completa:\n#{transcription}")
-        message = %{"chunks" => [%{"text" => transcription}]}
-        Phoenix.PubSub.broadcast(Whisper.PubSub, "transcription", {:transcription, Jason.encode!(message)})
-        File.rm!(path)
-      end)
-    end
 
     {:noreply, socket}
   end
 
+
   defp socket_id(socket), do: socket.transport_pid |> :erlang.pid_to_list() |> List.to_string()
   
-  def save_raw(ref, bin) do
-    File.mkdir_p!("recordings/")
-    filename = "#{ref}_#{Whisper.Counter.next(ref)}.raw"
-    path = Path.join("recordings", filename)
-    File.write!(path, bin)
-    {:ok, path}
-  end
 end

whisper/lib/whisper_web/components/layouts/app.html.heex

@@ -1,32 +1,6 @@
-<header class="px-4 sm:px-6 lg:px-8">
-  <div class="flex items-center justify-between border-b border-zinc-100 py-3 text-sm">
-    <div class="flex items-center gap-4">
-      <a href="/">
-        <img src={~p"/images/logo.svg"} width="36" />
-      </a>
-      <p class="bg-brand/5 text-brand rounded-full px-2 font-medium leading-6">
-        v{Application.spec(:phoenix, :vsn)}
-      </p>
-    </div>
-    <div class="flex items-center gap-4 font-semibold leading-6 text-zinc-900">
-      <a href="https://twitter.com/elixirphoenix" class="hover:text-zinc-700">
-        @elixirphoenix
-      </a>
-      <a href="https://github.com/phoenixframework/phoenix" class="hover:text-zinc-700">
-        GitHub
-      </a>
-      <a
-        href="https://hexdocs.pm/phoenix/overview.html"
-        class="rounded-lg bg-zinc-100 px-2 py-1 hover:bg-zinc-200/80"
-      >
-        Get Started <span aria-hidden="true">&rarr;</span>
-      </a>
-    </div>
-  </div>
-</header>
+
 <main class="px-4 py-20 sm:px-6 lg:px-8">
   <div class="mx-auto max-w-2xl">
-    <.flash_group flash={@flash} />
     {@inner_content}
   </div>
 </main>

whisper/lib/whisper_web/live/vad_live.ex

@@ -9,6 +9,7 @@ defmodule WhisperWeb.VadLive do
             socket
             |> assign(:transcription, "")
             |> assign(:started, false)
+            |> assign(:transcriptions, [])
 
         {:ok, socket}
     end
@@ -18,31 +19,50 @@ defmodule WhisperWeb.VadLive do
         {:noreply, assign(socket, started: true)}
     end
 
-    def handle_info({:transcription, raw_json}, socket) do
-        new_text =
-            raw_json
-            |> Jason.decode!()
-            |> get_in(["chunks", Access.at(0), "text"])
-        {:noreply, update(socket, :transcription, &(&1 <> " " <> new_text))}
+    def handle_event("stop_vad", _params, socket) do
+        push_event(socket, "stop-vad", %{})
+        {:noreply, assign(socket, started: false)}
     end
 
+    def handle_info({:transcription, %{"received_at" => ts, "text" => new_text}}, socket) do
+        updated_transcriptions =
+            [%{received_at: ts, text: new_text} | socket.assigns.transcriptions]
+            |> Enum.sort_by(& &1.received_at)
+
+        final_text = 
+            updated_transcriptions
+            |> Enum.map_join(" ", & &1.text)
+
+        socket =
+            socket
+            |> assign(:transcriptions, updated_transcriptions)
+            |> assign(:transcription, final_text)
+
+        {:noreply, socket}
+    end
+
+    
     def render(assigns) do
-        ~H"""
+    ~H"""
         <div id="vad-container" phx-hook="VadHook">
-            <button phx-click="start_vad" class="btn btn-primary">🎙 Iniciar VAD</button>
+            <%= if !@started do %>
+                <button phx-click="start_vad" class="btn btn-primary">🎙 Iniciar VAD</button>
+            <% end %>
 
             <div id="vad-status" class="mt-4 text-sm text-gray-700"></div>
         </div>
-    
+
         <div id="transcriptionContainer" class="w-full max-w-2xl space-y-4">
             <%= if @transcription != "" do %>
-                <div class="p-4 bg-gray-100 rounded shadow-md">
-                    <h2 class="text-sm font-semibold text-gray-700 mb-2">✅ Transcripción</h2>
-                    <p class="text-green-600 whitespace-pre-wrap break-words text-sm leading-relaxed"><%= @transcription %></p>
-                </div>
+            <div class="p-4 bg-gray-100 rounded shadow-md">
+                <h2 class="text-sm font-semibold text-gray-700 mb-2">✅ Transcripción</h2>
+                <p class="text-green-600 whitespace-pre-wrap break-words text-sm leading-relaxed">
+                <%= @transcription %>
+                </p>
+            </div>
             <% end %>
         </div>
-        """
+    """
     end
 
 end