speaker-detector 0.1.4__py3-none-any.whl → 0.1.6__py3-none-any.whl

speaker_detector/web/static/scripts/script.js

@@ -0,0 +1,22 @@
+import { setupAccordionNav } from "/static/components/accordion-nav/accordion-nav.js";
+import { setupMicTest } from "/static/components/mic-test/mic-test.js";
+import { setupEnrollSpeaker } from "/static/components/enroll-speaker/enroll-speaker.js";
+import { setupIdentifySpeaker } from "/static/components/identify-speaker/identify-speaker.js";
+import { setupMeetingMode } from "/static/components/meeting-mode/meeting-mode.js";
+import { setupRecordingsTab } from "/static/components/recordings-tab/recordings-tab.js";
+import { setupSpeakersList } from "/static/components/speakers-list/speakers-list.js";
+import { setupCorrection } from "/static/components/correction/correction.js";
+import { setupMicPopup } from "/static/components/mic-popup/mic-popup.js";
+
+// ✅ Export setup block so loader can run it later
+export function runSetup() {
+  setupAccordionNav();
+  setupMicTest();
+  setupEnrollSpeaker();
+  setupIdentifySpeaker();
+  setupMeetingMode();
+  setupRecordingsTab();
+
+  setupCorrection();
+  setupMicPopup();
+}

speaker_detector/web/static/style.css

@@ -0,0 +1,133 @@
+body {
+  font-family: sans-serif;
+  margin: 0;
+  padding: 0;
+  background: #f5f5f5;
+}
+
+.header {
+  background: #007acc;
+  color: white;
+  padding: 1rem;
+  text-align: center;
+}
+
+.workflow-accordion {
+  display: flex;
+  gap: 10px;
+  padding: 1rem;
+  background: white;
+  border-bottom: 1px solid #ddd;
+  flex-wrap: wrap;
+}
+
+.accordion-step {
+  flex: 1;
+  background: #e0e0e0;
+  padding: 10px;
+  border-radius: 8px;
+  text-align: center;
+  cursor: pointer;
+  transition: background 0.2s ease, color 0.2s ease;
+}
+
+.accordion-step.active {
+  background: #007acc;
+  color: white;
+}
+
+.accordion-content {
+  display: none;
+  padding: 1rem;
+  background: white;
+  border-bottom: 1px solid #ddd;
+}
+
+.accordion-content.active {
+  display: block;
+}
+
+button {
+  padding: 0.5rem 1rem;
+  margin-top: 0.5rem;
+  background: #007acc;
+  color: white;
+  border: none;
+  border-radius: 5px;
+  cursor: pointer;
+  transition: background 0.2s ease;
+}
+
+button:hover {
+  background: #005fa3;
+}
+
+.status {
+  margin-top: 0.5rem;
+  font-weight: bold;
+  color: #555;
+}
+
+canvas {
+  width: 100%;
+  height: 100px;
+  background: #111;
+  margin: 0.5rem 0;
+}
+
+ul {
+  list-style: none;
+  padding: 0;
+}
+
+ul li {
+  padding: 0.5rem 0;
+  border-bottom: 1px solid #eee;
+}
+
+input[type="text"] {
+  padding: 0.5rem;
+  width: 60%;
+  margin-right: 0.5rem;
+  border: 1px solid #ccc;
+  border-radius: 5px;
+}
+
+#timeline {
+  margin-top: 1rem;
+}
+
+#identify-result {
+  margin: 0.5rem 0;
+}
+
+@media (max-width: 768px) {
+  .workflow-accordion {
+    flex-direction: column;
+  }
+
+  .accordion-step {
+    flex: unset;
+  }
+}
+
+#loading-overlay {
+  position: fixed;
+  inset: 0;
+  background: #111;
+  color: white;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  font-size: 1.2rem;
+  z-index: 9999;
+}
+
+.loader {
+  animation: pulse 1.2s infinite ease-in-out;
+}
+
+@keyframes pulse {
+  0%, 100% { opacity: 0.4; }
+  50% { opacity: 1; }
+}

{speaker_detector-0.1.4.dist-info → speaker_detector-0.1.6.dist-info}/METADATA

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: speaker-detector
-Version: 0.1.4
-Summary: A CLI tool for speaker enrollment and identification using SpeechBrain.
+Version: 0.1.6
+Summary: A CLI + Web tool for speaker enrollment and identification using SpeechBrain.
 Author-email: Lara Whybrow <lara.whybrow@gmail.com>
 License: MIT
 Project-URL: Homepage, https://github.com/P0llen/speaker-detector
@@ -22,7 +22,14 @@ Description-Content-Type: text/markdown
 Requires-Dist: torch
 Requires-Dist: torchaudio
 Requires-Dist: speechbrain
-Requires-Dist: onnx
+Requires-Dist: flask
+Requires-Dist: flask-cors
+Requires-Dist: numpy
+Requires-Dist: sounddevice
+Requires-Dist: soundfile
+Requires-Dist: pydub
+
+Note: Still in development, as I am configuring the system for the most performant approach. Feel free to jump on the project with me. 
 
 # speaker-detector 🎙️
 
@@ -45,6 +52,12 @@ Install from [TestPyPI](https://test.pypi.org/):
 
 ```bash
 pip install --index-url https://test.pypi.org/simple/ speaker-detector
+
+When installing packages with a stale requirement file you might need to use:  pip install --break-system-packages soundfile to install on WSL Ubuntu
+
+Run this version with -m module flag if you are contributing and want to run server.py:
+python3 -m speaker_detector.server
+
 ```
 
 ## 🚀 Usage
@@ -99,3 +112,15 @@ onnxruntime
 
 
 NB: When pushing to Github, do not include any .identifier files.
+
+You can manually clean up stale embeddings that don’t match any existing speaker folder with a quick script:
+
+# Run inside your project root
+cd storage/embeddings
+for f in *.pt; do
+  speaker="${f%.pt}"
+  if [ ! -d "../speakers/$speaker" ]; then
+    echo "Deleting stale embedding: $f"
+    rm "$f"
+  fi
+done

speaker_detector-0.1.6.dist-info/RECORD

@@ -0,0 +1,25 @@
+speaker_detector/__main__.py,sha256=EClCwCzb6h6YBpt0hrnG4h0mlNhNePyg_xBNNSVm1os,65
+speaker_detector/cli.py,sha256=TxJhu3Pjhg41tkcu--aLtn0vZwBYyoVEef10zqSBzig,2619
+speaker_detector/core.py,sha256=zct1lNfeGB6Y7WB1rqNVPKyMKXu2g0KCuMkpAiyTLi0,5267
+speaker_detector/server copy.py,sha256=A1WplNK8yGe9AnEjrSRqHO-uJJsMwIlvEzDhgu72XyY,10723
+speaker_detector/server.py,sha256=6AkZ2drrE-jfV3M-4jdzobontsVxrGSrmQPykpSWyeo,3516
+speaker_detector/state.py,sha256=ikcTNsjBzbFNMWfpfDaWkWMOb01Y9wlzvCAu7ifQ30w,2761
+speaker_detector/model/ECAPA_TDNN.py,sha256=KB5T-ye4c9ZWgTgn_SMH-T_-qYSEHQJJtf3xHjsfNPk,19024
+speaker_detector/model/classifier.ckpt,sha256=_Z42NP5ovQpCfJXjVMDGdzdPYrP0NORbeFmZUNhg1TU,5534328
+speaker_detector/model/embedding_model.ckpt,sha256=BXXLZIRea5oQ25vLdNWsMrMmuNyQNSZx00Xi7j0BJqI,83316686
+speaker_detector/model/hyperparams.yaml,sha256=b3iFT6BLpZ52FDe3aiV106ul5QFt4-m2nwyaUHf7GkE,1919
+speaker_detector/model/label_encoder.ckpt,sha256=4Tw6Fnu0ESaFZw7oltIOK1Za8Ws6TO6qhon6TSKtuLk,128619
+speaker_detector/model/mean_var_norm_emb.ckpt,sha256=zXAiWwWze-ZPxaleJDldgEIx1D90suHlpRPbe2mzTDM,1921
+speaker_detector/web/static/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+speaker_detector/web/static/favicon.ico,sha256=JzURAUfKDB96EFxNo00a0pAG0DQWY8v7YDiBJeRDn9g,496
+speaker_detector/web/static/index.html,sha256=0Hmfzmyjqm_MyFyoNJY51EzLHFFamKXlFygGDy2TDX0,1218
+speaker_detector/web/static/style.css,sha256=R5pGvJqh5j95o6RSmpEFDNQMxNM8yQ7ie7Hlms1HZ7U,2025
+speaker_detector/web/static/scripts/loader copy.js,sha256=BwhTS_ulxb62cwF6qAk1ng-bMBM-29l1rQ7W7QFTYPk,333
+speaker_detector/web/static/scripts/loader.js,sha256=OWgmKfZ0E7bKaVPR47Q5aA-JKFeD3741K_wS_lqphz4,503
+speaker_detector/web/static/scripts/script copy.js,sha256=LLcKKjTjXEy9yj5e4gQhgIsItRLYAJe_9V7mppyM8Bc,31494
+speaker_detector/web/static/scripts/script.js,sha256=UyHWk1HrkWW6ZyxQYRLOPbtCMLHcBbNudMfTJ9k3neA,1023
+speaker_detector-0.1.6.dist-info/METADATA,sha256=WSOiZtAgOwze7mJFPETyqg7VtTOOHp4nQewa-GHUAlc,5414
+speaker_detector-0.1.6.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+speaker_detector-0.1.6.dist-info/entry_points.txt,sha256=2B30ee2cTyeeA49x_TBURl53bDRiLWGK3NWhb9rlK3s,63
+speaker_detector-0.1.6.dist-info/top_level.txt,sha256=PJ5rfvd3GAbzMbc7-Fwhtufjf6HxzzTiiHociOy7RiM,17
+speaker_detector-0.1.6.dist-info/RECORD,,

{speaker_detector-0.1.4.dist-info → speaker_detector-0.1.6.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (80.8.0)
+Generator: setuptools (80.9.0)
 Root-Is-Purelib: true
 Tag: py3-none-any
 

speaker_detector/analyze.py

@@ -1,59 +0,0 @@
-from pathlib import Path
-import torchaudio
-import torch
-from speaker_detector.core import get_embedding, STORAGE_DIR
-
-CHUNK_DURATION = 2.5  # seconds
-
-def match_speaker(embedding, speaker_embeddings):
-    scores = {}
-    for name, emb in speaker_embeddings.items():
-        score = torch.nn.functional.cosine_similarity(emb, embedding, dim=0).item()
-        scores[name] = score
-    if not scores:
-        return "unknown", 0.0
-    best = max(scores.items(), key=lambda kv: kv[1])
-    return best[0], round(best[1], 3)
-
-def analyze_meeting(wav_path):
-    waveform, sample_rate = torchaudio.load(wav_path)
-    duration_sec = waveform.shape[1] / sample_rate
-
-    chunk_samples = int(CHUNK_DURATION * sample_rate)
-    num_chunks = int(waveform.shape[1] / chunk_samples)
-
-    # Load enrolled speaker embeddings
-    speaker_embeddings = {}
-    for spk_dir in STORAGE_DIR.iterdir():
-        if not spk_dir.is_dir():
-            continue
-        wavs = list(spk_dir.glob("*.wav"))
-        if not wavs:
-            continue
-        # Average multiple embeddings
-        embs = [get_embedding(str(wav)) for wav in wavs]
-        speaker_embeddings[spk_dir.name] = torch.stack(embs).mean(dim=0)
-
-    results = []
-
-    for i in range(num_chunks):
-        start_sample = i * chunk_samples
-        end_sample = start_sample + chunk_samples
-        chunk = waveform[:, start_sample:end_sample]
-
-        tmp_path = Path(wav_path).parent / f"tmp_chunk_{i}.wav"
-        torchaudio.save(str(tmp_path), chunk, sample_rate)
-
-        embedding = get_embedding(str(tmp_path))
-        speaker, score = match_speaker(embedding, speaker_embeddings)
-
-        results.append({
-            "start": round(i * CHUNK_DURATION, 2),
-            "end": round((i + 1) * CHUNK_DURATION, 2),
-            "speaker": speaker,
-            "score": score
-        })
-
-        tmp_path.unlink()  # clean up
-
-    return results

speaker_detector/combine.py

@@ -1,22 +0,0 @@
-import torch
-import os
-
-def combine_embeddings_from_folder(folder_path, output_path):
-    speaker_data = {}
-
-    for fname in os.listdir(folder_path):
-        if fname.endswith(".pt"):
-            label = os.path.splitext(fname)[0]
-            fpath = os.path.join(folder_path, fname)
-            tensor = torch.load(fpath, map_location="cpu")
-            if not isinstance(tensor, torch.Tensor):
-                print(f"❌ Skipping {fname}: not a valid tensor")
-                continue
-            speaker_data[label] = tensor
-
-    if not speaker_data:
-        print("⚠️ No valid .pt files found.")
-        return
-
-    torch.save(speaker_data, output_path)
-    print(f"✅ Combined {len(speaker_data)} speakers into {output_path}")

speaker_detector/export_embeddings.py

@@ -1,41 +0,0 @@
-import torch
-import json
-
-def export_embeddings_to_json(pt_path, json_path):
-    """
-    Converts a .pt file containing speaker embeddings into a
-    JSON file for use in the browser frontend.
-
-    Expected input format:
-        {
-            "lara": tensor([...]),
-            "guest": tensor([...]),
-            ...
-        }
-
-    Output format:
-        [
-            { "label": "lara", "vector": [...] },
-            { "label": "guest", "vector": [...] },
-            ...
-        ]
-    """
-    data = torch.load(pt_path, map_location="cpu")
-
-    if not isinstance(data, dict):
-        raise ValueError("Expected a dict of {label: tensor} in the .pt file")
-
-    converted = []
-    for label, tensor in data.items():
-        if not isinstance(tensor, torch.Tensor):
-            print(f"⚠️ Skipping {label}: not a tensor")
-            continue
-        converted.append({
-            "label": label,
-            "vector": tensor.tolist()
-        })
-
-    with open(json_path, "w") as f:
-        json.dump(converted, f, indent=2)
-
-    print(f"✅ Exported {len(converted)} speaker embeddings to {json_path}")

speaker_detector/export_model.py

@@ -1,40 +0,0 @@
-import torch
-from speechbrain.lobes.models.ECAPA_TDNN import ECAPA_TDNN
-from collections import OrderedDict
-
-def export_model_to_onnx(ckpt_path, out_path):
-    model = ECAPA_TDNN(
-        input_size=80,
-        channels=[1024, 1024, 1024, 1024, 3072],
-        kernel_sizes=[5, 3, 3, 3, 1],
-        dilations=[1, 2, 3, 4, 1],
-        attention_channels=128,
-        lin_neurons=192,
-    )
-
-    state_dict = torch.load(ckpt_path, map_location="cpu")
-
-    if "model" in state_dict:
-        state_dict = state_dict["model"]
-
-    new_state_dict = OrderedDict()
-    for k, v in state_dict.items():
-        if k.startswith("embedding_model."):
-            k = k[len("embedding_model."):]
-        new_state_dict[k] = v
-
-    model.load_state_dict(new_state_dict)
-    model.eval()
-
-    dummy_input = torch.randn(1, 200, 80)
-    torch.onnx.export(
-        model,
-        dummy_input,
-        out_path,
-        input_names=["features"],
-        output_names=["embedding"],
-        dynamic_axes={"features": {0: "batch", 1: "time"}},
-        opset_version=12,
-    )
-
-    print(f"✅ Exported ECAPA-TDNN to {out_path}")

speaker_detector/generate_summary.py

@@ -1,110 +0,0 @@
-import os
-import torch
-import torchaudio
-import requests
-from pathlib import Path
-from pydub import AudioSegment
-from dotenv import load_dotenv
-from speaker_detector.core import get_embedding, STORAGE_DIR
-
-load_dotenv()
-OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
-
-CHUNK_DURATION = 8  # seconds
-SCORE_THRESHOLD = 0.6
-MIN_VALID_DURATION = 1.0  # seconds
-WHISPER_API_URL = "https://api.openai.com/v1/audio/transcriptions"
-
-def match_speaker(embedding, speaker_embeddings):
-    scores = {
-        name: torch.nn.functional.cosine_similarity(emb, embedding, dim=0).item()
-        for name, emb in speaker_embeddings.items()
-    }
-    if not scores:
-        return "unknown", 0.0
-    best = max(scores.items(), key=lambda kv: kv[1])
-    return best[0], round(best[1], 3)
-
-def transcribe_full_audio(wav_path: Path) -> str:
-    try:
-        with open(wav_path, "rb") as f:
-            response = requests.post(
-                WHISPER_API_URL,
-                headers={"Authorization": f"Bearer {OPENAI_API_KEY}"},
-                files={"file": (wav_path.name, f, "audio/wav")},
-                data={
-                    "model": "whisper-1",
-                    "response_format": "json",
-                    "temperature": 0.2,
-                    "language": "en",
-                    "prompt": "This is a meeting transcription.",
-                },
-                timeout=120
-            )
-        response.raise_for_status()
-        return response.json()["text"].strip()
-    except Exception as e:
-        print(f"❌ Whisper failed: {e}")
-        return ""
-
-def is_valid_audio(path):
-    try:
-        waveform, sample_rate = torchaudio.load(str(path))
-        duration_sec = waveform.shape[1] / sample_rate
-        return duration_sec >= MIN_VALID_DURATION
-    except Exception:
-        return False
-
-def generate_summary(meeting_dir: Path):
-    meeting_dir = meeting_dir.resolve()
-    chunk_files = sorted([
-        f for f in meeting_dir.iterdir()
-        if f.name.startswith("chunk_") and f.suffix == ".wav" and is_valid_audio(f)
-    ])
-
-    if not chunk_files:
-        return {"warning": "No valid .wav chunks found in meeting folder.", "segments": []}
-
-    # Merge all chunks into one file
-    combined = AudioSegment.empty()
-    for f in chunk_files:
-        combined += AudioSegment.from_wav(f)
-    merged_path = meeting_dir / "combined.wav"
-    combined.export(merged_path, format="wav")
-
-    # Get full transcript
-    full_text = transcribe_full_audio(merged_path)
-    print("🧠 Full transcript:", full_text)
-
-    # Load speaker embeddings
-    speaker_embeddings = {}
-    for spk_dir in STORAGE_DIR.iterdir():
-        if spk_dir.is_dir():
-            wavs = [w for w in spk_dir.glob("*.wav") if is_valid_audio(w)]
-            if wavs:
-                embs = [get_embedding(str(w)) for w in wavs]
-                speaker_embeddings[spk_dir.name] = torch.stack(embs).mean(dim=0)
-
-    segments = []
-    total = len(chunk_files)
-
-    for idx, chunk in enumerate(chunk_files):
-        try:
-            emb = get_embedding(chunk)
-            speaker, score = match_speaker(emb, speaker_embeddings)
-            segment_text = f"[chunk {idx+1}]"
-            segments.append({
-                "timestamp": idx * CHUNK_DURATION,
-                "speaker": speaker if score >= SCORE_THRESHOLD else "unknown",
-                "score": round(score, 3),
-                "text": segment_text,
-                "progress": round((idx + 1) / total * 100)
-            })
-        except Exception as e:
-            print(f"❌ Failed on {chunk.name}: {e}")
-
-    return {
-        "transcript": full_text,
-        "segments": segments if segments else [],
-        "warning": None if segments else "No speaker segments found."
-    }

speaker_detector-0.1.4.dist-info/RECORD

@@ -1,15 +0,0 @@
-speaker_detector/ECAPA_TDNN.py,sha256=KB5T-ye4c9ZWgTgn_SMH-T_-qYSEHQJJtf3xHjsfNPk,19024
-speaker_detector/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-speaker_detector/__main__.py,sha256=EClCwCzb6h6YBpt0hrnG4h0mlNhNePyg_xBNNSVm1os,65
-speaker_detector/analyze.py,sha256=sA8qyzczdHUbJw2_1JIbXn1WpiKC5dHLPRtPPoppJzY,1943
-speaker_detector/cli.py,sha256=TKci4o4Fru-3NqUkPDRQRvtis2niNEAh9sQWwE5t6Us,3521
-speaker_detector/combine.py,sha256=yCiqG6VMojz0CxSTPqjx0RrUban8oFIcKlA1zFMzaU4,761
-speaker_detector/core.py,sha256=lQNOcmZs2IJOqrNKlk1BeVQX6tzc7BSpeP5Gordff-E,3586
-speaker_detector/export_embeddings.py,sha256=OxNXadzEiMEJgpmCG6HHFncUX7DumFvTOys1R6UMUnw,1151
-speaker_detector/export_model.py,sha256=qVVT2wSCnsPA8pSAEEyIMkY7Kc8uAgepc03MxBMT3xU,1146
-speaker_detector/generate_summary.py,sha256=oTWEf2bxTCRIUl8L17-J64FyhRbCPnDjihFluEnBWc8,3726
-speaker_detector-0.1.4.dist-info/METADATA,sha256=c0l2KJOvthuvSJ_g7QLb23sYDHqXhm4wi8dJYChEB8M,4564
-speaker_detector-0.1.4.dist-info/WHEEL,sha256=zaaOINJESkSfm_4HQVc5ssNzHCPXhJm0kEUakpsEHaU,91
-speaker_detector-0.1.4.dist-info/entry_points.txt,sha256=2B30ee2cTyeeA49x_TBURl53bDRiLWGK3NWhb9rlK3s,63
-speaker_detector-0.1.4.dist-info/top_level.txt,sha256=PJ5rfvd3GAbzMbc7-Fwhtufjf6HxzzTiiHociOy7RiM,17
-speaker_detector-0.1.4.dist-info/RECORD,,