whisper-large-v3

Running

App Files Files Community

staraks commited on Nov 18

Commit

9bf49ce

verified ·

1 Parent(s): f0f2431

Update app.py

Browse files

Files changed (1) hide show

app.py +232 -7

app.py CHANGED Viewed

@@ -186,7 +186,6 @@ def extract_zip_to_temp(zip_file: UploadFile, password: Optional[str]) -> List[s
             detail=f"Failed to open ZIP file. Check password / integrity. {e}",
         )
-    # Only top-level; nested dirs can be added if needed.
     files = [os.path.join(outdir, f) for f in os.listdir(outdir)]
     return files
@@ -223,6 +222,35 @@ def health():
     return "OK"
 # ---------- 1. Multi-file transcription (JSON) ----------
 @app.post("/api/transcribe/files", response_model=TranscriptionResponse)
@@ -505,6 +533,56 @@ HTML_UI = """
       margin-top: -4px;
       margin-bottom: 8px;
     }
     @media (max-width: 768px) {
       header {
         padding: 12px 16px;
@@ -518,7 +596,11 @@ HTML_UI = """
 <body>
   <header>
     <h1>Whisper Large V3 – Medical Batch Transcription</h1>
-    <p>Upload multiple audio files or a password-protected ZIP. Mode: general or medical_en. API docs at <code>/docs</code>.</p>
   </header>
   <main>
     <div class="card">
@@ -528,7 +610,9 @@ HTML_UI = """
           <h3>Inputs</h3>
           <label for="files_input">Audio files</label>
           <input id="files_input" type="file" multiple accept="audio/*" />
-          <div class="small-hint">You can select multiple audio files.</div>
           <label for="files_mode">Mode</label>
           <select id="files_mode">
@@ -578,10 +662,119 @@ HTML_UI = """
       </div>
     </div>
     <div id="status"></div>
   </main>
   <script>
     async function postForm(url, formData, expectBlob = false) {
       const res = await fetch(url, {
         method: "POST",
@@ -608,6 +801,27 @@ HTML_UI = """
       document.getElementById("status").innerText = text || "";
     }
     // ------- Multi-files JSON -------
     document.getElementById("btn_files_json").addEventListener("click", async () => {
       const filesInput = document.getElementById("files_input");
@@ -627,15 +841,17 @@ HTML_UI = """
       setStatus("Transcribing multiple files… (this may take some time for large audio)");
       out.value = "";
       try {
         const data = await postForm("/api/transcribe/files", formData, false);
         out.value = data.combined_transcript || "";
         setStatus("Done.");
       } catch (err) {
         console.error(err);
         alert(err.message);
-        setStatus("Error during transcription.");
       }
     });
@@ -656,6 +872,7 @@ HTML_UI = """
       formData.append("mode", mode);
       setStatus("Generating DOCX for multi-file transcription…");
       try {
         const blob = await postForm("/api/transcribe/files/docx", formData, true);
@@ -668,10 +885,11 @@ HTML_UI = """
         a.remove();
         window.URL.revokeObjectURL(url);
         setStatus("DOCX downloaded.");
       } catch (err) {
         console.error(err);
         alert(err.message);
-        setStatus("Error during DOCX generation.");
       }
     });
@@ -694,15 +912,17 @@ HTML_UI = """
       setStatus("Transcribing ZIP contents…");
       out.value = "";
       try {
         const data = await postForm("/api/transcribe/zip", formData, false);
         out.value = data.combined_transcript || "";
         setStatus("Done.");
       } catch (err) {
         console.error(err);
         alert(err.message);
-        setStatus("Error during ZIP transcription.");
       }
     });
@@ -723,6 +943,7 @@ HTML_UI = """
       formData.append("mode", mode);
       setStatus("Generating DOCX from ZIP contents…");
       try {
         const blob = await postForm("/api/transcribe/zip/docx", formData, true);
@@ -735,12 +956,16 @@ HTML_UI = """
         a.remove();
         window.URL.revokeObjectURL(url);
         setStatus("DOCX downloaded.");
       } catch (err) {
         console.error(err);
         alert(err.message);
-        setStatus("Error during ZIP DOCX generation.");
       }
     });
   </script>
 </body>
 </html>

             detail=f"Failed to open ZIP file. Check password / integrity. {e}",
         )
     files = [os.path.join(outdir, f) for f in os.listdir(outdir)]
     return files
     return "OK"
+@app.get("/self-test")
+def self_test():
+    """
+    Basic self-check:
+    - can we create/load the pipeline?
+    - what device are we using?
+    """
+    try:
+        pipe = get_pipeline()
+        model_name = getattr(pipe.model, "name_or_path", MODEL_NAME)
+        dev = "cuda" if device == 0 else str(device)
+        return JSONResponse(
+            {
+                "status": "ok",
+                "message": "Pipeline loaded successfully.",
+                "model": model_name,
+                "device": dev,
+            }
+        )
+    except Exception as e:
+        return JSONResponse(
+            {
+                "status": "error",
+                "message": f"Pipeline failed to load: {e}",
+            },
+            status_code=500,
+        )
 # ---------- 1. Multi-file transcription (JSON) ----------
 @app.post("/api/transcribe/files", response_model=TranscriptionResponse)
       margin-top: -4px;
       margin-bottom: 8px;
     }
+    code {
+      font-family: ui-monospace, SFMono-Regular, Menlo, Monaco, Consolas, "Liberation Mono", "Courier New", monospace;
+      font-size: 12px;
+    }
+    pre {
+      background: #0b1120;
+      color: #e5e7eb;
+      padding: 10px 12px;
+      border-radius: 10px;
+      overflow-x: auto;
+      font-size: 12px;
+      line-height: 1.5;
+    }
+    a {
+      color: #1d4ed8;
+      text-decoration: none;
+    }
+    a:hover {
+      text-decoration: underline;
+    }
+    /* Progress bar */
+    #progress-wrapper {
+      margin-top: 10px;
+      margin-bottom: 4px;
+      font-size: 12px;
+      color: #4b5563;
+    }
+    #progress-track {
+      width: 100%;
+      height: 8px;
+      background: #e5e7eb;
+      border-radius: 999px;
+      overflow: hidden;
+      margin-top: 4px;
+    }
+    #progress-fill {
+      height: 100%;
+      width: 0%;
+      background: #111827;
+      border-radius: 999px;
+      transition: width 0.2s ease-out;
+    }
+    #progress-text {
+      font-size: 11px;
+      color: #6b7280;
+      margin-top: 3px;
+      min-height: 14px;
+    }
     @media (max-width: 768px) {
       header {
         padding: 12px 16px;
 <body>
   <header>
     <h1>Whisper Large V3 – Medical Batch Transcription</h1>
+    <p>
+      Upload multiple audio files or a password-protected ZIP.
+      Mode: <code>general</code> or <code>medical_en</code>.
+      API docs at <code>/docs</code>.
+    </p>
   </header>
   <main>
     <div class="card">
           <h3>Inputs</h3>
           <label for="files_input">Audio files</label>
           <input id="files_input" type="file" multiple accept="audio/*" />
+          <div class="small-hint">
+            You can select multiple audio files.
+          </div>
           <label for="files_mode">Mode</label>
           <select id="files_mode">
       </div>
     </div>
+    <!-- 3. Quick examples -->
+    <div class="card">
+      <h2>3. Quick examples <span class="pill">API & sample audio</span></h2>
+      <h3>Sample audio for testing (download & upload above)</h3>
+      <p class="small-hint">
+        1. Download this small public sample file<br>
+        2. Upload it in section 1 and click <strong>Transcribe → JSON</strong>
+      </p>
+      <p>
+        👉 <a href="https://huggingface.co/datasets/Narsil/asr_dummy/resolve/main/mlk.flac" target="_blank" rel="noopener">
+          Download example audio (mlk.flac)
+        </a>
+      </p>
+      <h3>Example: cURL for multi-file JSON</h3>
+      <p class="small-hint">Replace <code>@path/to/audio1.flac</code> with your local file path.</p>
+      <pre><code>curl -X POST \\
+  "https://staraks-whisper-large-v3.hf.space/api/transcribe/files" \\
+  -H "Accept: application/json" \\
+  -F "mode=medical_en" \\
+  -F "files=@path/to/audio1.flac" \\
+  -F "files=@path/to/audio2.wav"</code></pre>
+      <h3>Example: cURL for ZIP JSON</h3>
+      <p class="small-hint">ZIP file contains multiple audio files. Password field is optional.</p>
+      <pre><code>curl -X POST \\
+  "https://staraks-whisper-large-v3.hf.space/api/transcribe/zip" \\
+  -H "Accept: application/json" \\
+  -F "mode=medical_en" \\
+  -F "file=@path/to/audios.zip" \\
+  -F "password="</code></pre>
+    </div>
+    <!-- 4. System self-check -->
+    <div class="card">
+      <h2>4. System self-check <span class="pill">Model & API status</span></h2>
+      <p class="small-hint">
+        Use this to quickly verify that the API is running and the Whisper pipeline can be loaded.
+      </p>
+      <button class="btn-primary" id="btn_self_test">Run self-test</button>
+      <pre id="self_test_output"><code>Click "Run self-test" to see status...</code></pre>
+    </div>
+    <!-- Progress & status -->
+    <div id="progress-wrapper">
+      <div>Transcription progress</div>
+      <div id="progress-track">
+        <div id="progress-fill"></div>
+      </div>
+      <div id="progress-text">Idle</div>
+    </div>
     <div id="status"></div>
   </main>
   <script>
+    let __progressTimer = null;
+    let __progressValue = 0;
+    function setProgress(value, label) {
+      __progressValue = Math.max(0, Math.min(100, value));
+      const fill = document.getElementById("progress-fill");
+      const text = document.getElementById("progress-text");
+      if (fill) {
+        fill.style.width = __progressValue + "%";
+      }
+      if (text) {
+        text.innerText = label + " (" + __progressValue.toFixed(0) + "%)";
+      }
+    }
+    function resetProgress() {
+      if (__progressTimer) {
+        clearInterval(__progressTimer);
+        __progressTimer = null;
+      }
+      setProgress(0, "Idle");
+    }
+    function startSimulatedProgress(label) {
+      if (__progressTimer) {
+        clearInterval(__progressTimer);
+      }
+      let p = 5;
+      setProgress(p, label);
+      __progressTimer = setInterval(() => {
+        if (p < 90) {
+          p += Math.random() * 10;
+          if (p > 90) p = 90;
+          setProgress(p, label);
+        }
+      }, 600);
+    }
+    function finishProgress(label) {
+      if (__progressTimer) {
+        clearInterval(__progressTimer);
+        __progressTimer = null;
+      }
+      setProgress(100, label);
+      setTimeout(() => {
+        resetProgress();
+      }, 2000);
+    }
+    function errorProgress(message) {
+      if (__progressTimer) {
+        clearInterval(__progressTimer);
+        __progressTimer = null;
+      }
+      setProgress(0, "Error");
+      setStatus(message);
+    }
     async function postForm(url, formData, expectBlob = false) {
       const res = await fetch(url, {
         method: "POST",
       document.getElementById("status").innerText = text || "";
     }
+    // ------- Self test -------
+    document.getElementById("btn_self_test").addEventListener("click", async () => {
+      const out = document.getElementById("self_test_output");
+      out.textContent = "Running self-test...";
+      setStatus("Running self-test…");
+      try {
+        const res = await fetch("/self-test");
+        const data = await res.json();
+        out.textContent = JSON.stringify(data, null, 2);
+        if (data.status === "ok") {
+          setStatus("Self-test OK – model and API are working.");
+        } else {
+          setStatus("Self-test reported an error.");
+        }
+      } catch (err) {
+        console.error(err);
+        out.textContent = "Self-test failed: " + err.message;
+        setStatus("Self-test failed.");
+      }
+    });
     // ------- Multi-files JSON -------
     document.getElementById("btn_files_json").addEventListener("click", async () => {
       const filesInput = document.getElementById("files_input");
       setStatus("Transcribing multiple files… (this may take some time for large audio)");
       out.value = "";
+      startSimulatedProgress("Transcribing files");
       try {
         const data = await postForm("/api/transcribe/files", formData, false);
         out.value = data.combined_transcript || "";
         setStatus("Done.");
+        finishProgress("Transcription complete");
       } catch (err) {
         console.error(err);
         alert(err.message);
+        errorProgress("Error during transcription.");
       }
     });
       formData.append("mode", mode);
       setStatus("Generating DOCX for multi-file transcription…");
+      startSimulatedProgress("Generating DOCX");
       try {
         const blob = await postForm("/api/transcribe/files/docx", formData, true);
         a.remove();
         window.URL.revokeObjectURL(url);
         setStatus("DOCX downloaded.");
+        finishProgress("DOCX ready");
       } catch (err) {
         console.error(err);
         alert(err.message);
+        errorProgress("Error during DOCX generation.");
       }
     });
       setStatus("Transcribing ZIP contents…");
       out.value = "";
+      startSimulatedProgress("Transcribing ZIP");
       try {
         const data = await postForm("/api/transcribe/zip", formData, false);
         out.value = data.combined_transcript || "";
         setStatus("Done.");
+        finishProgress("ZIP transcription complete");
       } catch (err) {
         console.error(err);
         alert(err.message);
+        errorProgress("Error during ZIP transcription.");
       }
     });
       formData.append("mode", mode);
       setStatus("Generating DOCX from ZIP contents…");
+      startSimulatedProgress("Generating ZIP DOCX");
       try {
         const blob = await postForm("/api/transcribe/zip/docx", formData, true);
         a.remove();
         window.URL.revokeObjectURL(url);
         setStatus("DOCX downloaded.");
+        finishProgress("ZIP DOCX ready");
       } catch (err) {
         console.error(err);
         alert(err.message);
+        errorProgress("Error during ZIP DOCX generation.");
       }
     });
+    // Initial state
+    resetProgress();
   </script>
 </body>
 </html>