transcribe-studio 0.2.0__tar.gz

transcribe_studio-0.2.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Mishkat Quantum Labs
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

transcribe_studio-0.2.0/PKG-INFO

@@ -0,0 +1,128 @@
+Metadata-Version: 2.4
+Name: transcribe-studio
+Version: 0.2.0
+Summary: Local classroom audio transcription with projects, WER evaluation, and pluggable transcript formats
+Author: Mishkat Quantum Labs
+License-Expression: MIT
+Project-URL: Homepage, https://github.com/Mishkat-Quantum-Labs/transcribe-studio
+Project-URL: Repository, https://github.com/Mishkat-Quantum-Labs/transcribe-studio
+Project-URL: Issues, https://github.com/Mishkat-Quantum-Labs/transcribe-studio/issues
+Project-URL: Changelog, https://github.com/Mishkat-Quantum-Labs/transcribe-studio/blob/main/CHANGELOG.md
+Keywords: transcription,classroom,audio,wer,speech,evaluation
+Classifier: Development Status :: 4 - Beta
+Classifier: Environment :: Web Environment
+Classifier: Framework :: FastAPI
+Classifier: Intended Audience :: Science/Research
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Multimedia :: Sound/Audio :: Speech
+Classifier: Topic :: Scientific/Engineering
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: fastapi>=0.115.0
+Requires-Dist: uvicorn[standard]>=0.32.0
+Requires-Dist: jinja2>=3.1.0
+Requires-Dist: python-multipart>=0.0.12
+Requires-Dist: aiofiles>=24.1.0
+Provides-Extra: dev
+Requires-Dist: pytest>=8.0; extra == "dev"
+Requires-Dist: build>=1.0; extra == "dev"
+Requires-Dist: twine>=5.0; extra == "dev"
+Dynamic: license-file
+
+# Transcribe Studio
+
+[![CI](https://github.com/Mishkat-Quantum-Labs/transcribe-studio/actions/workflows/ci.yml/badge.svg)](https://github.com/Mishkat-Quantum-Labs/transcribe-studio/actions/workflows/ci.yml)
+
+A local, browser-based tool for classroom audio transcription. Organize work by **project**, split audio into timed chunks, label speakers in free text, and evaluate human transcripts against LLM output (WER + semantic WER).
+
+Built for researchers and annotators who need millisecond timestamps and exportable data — without Label Studio complexity.
+
+## Features
+
+- **Projects** — group recordings by class, session, or study
+- **Waveform editor** — divide audio into chunks (by duration or count), overlap speakers at the same timestamp
+- **Chunk playback** — play one chunk at a time with **speed up/down** (0.25×–2×, keys `,` / `.`)
+- **Exports** — TXT, Markdown, JSON, CSV, SRT, WebVTT
+- **LLM evaluation** — paste or upload hypothesis transcripts; strict + semantic WER
+- **Pluggable formats** — timestamp/speaker lines, JSON segments, plain text (TOML-driven)
+
+## Quick start
+
+### With uv (recommended)
+
+```bash
+git clone https://github.com/Mishkat-Quantum-Labs/transcribe-studio.git
+cd transcribe-studio
+uv venv
+uv pip install -e ".[dev]"
+uv run transcribe-studio
+```
+
+Open **http://127.0.0.1:8082**
+
+### With pip
+
+```bash
+pip install transcribe-studio
+transcribe-studio
+```
+
+## Usage
+
+1. Create a **project** from the dashboard
+2. **Upload** an MP3/WAV/M4A/OGG/FLAC recording
+3. **Divide** the wave into chunks, then transcribe each segment
+4. Use **Evaluation** to compare your transcript against an LLM upload
+5. **Export** when done
+
+Data is stored under `~/.transcribe-studio/` (override with `TRANSCRIBE_STUDIO_DATA`).
+
+## Development
+
+```bash
+uv pip install -e ".[dev]"
+uv run pytest
+```
+
+## Publishing
+
+### PyPI via uv (recommended)
+
+```bash
+uv build
+uv publish   # uses UV_PUBLISH_TOKEN or prompts for PyPI credentials
+```
+
+### PyPI via pip/twine
+
+```bash
+pip install build twine
+python -m build
+twine upload dist/*
+```
+
+### GitHub release
+
+```bash
+git tag v0.2.0
+git push origin v0.2.0
+gh release create v0.2.0 dist/*
+```
+
+## Configuration
+
+Evaluation and transcript import settings ship inside the package:
+
+- `app/config/evaluation.toml`
+- `app/config/transcript_formats.toml`
+- `app/config/languages/en.toml`
+
+## License
+
+MIT — see [LICENSE](LICENSE).
+
+## Contributing
+
+Issues and PRs welcome at [github.com/Mishkat-Quantum-Labs/transcribe-studio](https://github.com/Mishkat-Quantum-Labs/transcribe-studio).

transcribe_studio-0.2.0/README.md

@@ -0,0 +1,95 @@
+# Transcribe Studio
+
+[![CI](https://github.com/Mishkat-Quantum-Labs/transcribe-studio/actions/workflows/ci.yml/badge.svg)](https://github.com/Mishkat-Quantum-Labs/transcribe-studio/actions/workflows/ci.yml)
+
+A local, browser-based tool for classroom audio transcription. Organize work by **project**, split audio into timed chunks, label speakers in free text, and evaluate human transcripts against LLM output (WER + semantic WER).
+
+Built for researchers and annotators who need millisecond timestamps and exportable data — without Label Studio complexity.
+
+## Features
+
+- **Projects** — group recordings by class, session, or study
+- **Waveform editor** — divide audio into chunks (by duration or count), overlap speakers at the same timestamp
+- **Chunk playback** — play one chunk at a time with **speed up/down** (0.25×–2×, keys `,` / `.`)
+- **Exports** — TXT, Markdown, JSON, CSV, SRT, WebVTT
+- **LLM evaluation** — paste or upload hypothesis transcripts; strict + semantic WER
+- **Pluggable formats** — timestamp/speaker lines, JSON segments, plain text (TOML-driven)
+
+## Quick start
+
+### With uv (recommended)
+
+```bash
+git clone https://github.com/Mishkat-Quantum-Labs/transcribe-studio.git
+cd transcribe-studio
+uv venv
+uv pip install -e ".[dev]"
+uv run transcribe-studio
+```
+
+Open **http://127.0.0.1:8082**
+
+### With pip
+
+```bash
+pip install transcribe-studio
+transcribe-studio
+```
+
+## Usage
+
+1. Create a **project** from the dashboard
+2. **Upload** an MP3/WAV/M4A/OGG/FLAC recording
+3. **Divide** the wave into chunks, then transcribe each segment
+4. Use **Evaluation** to compare your transcript against an LLM upload
+5. **Export** when done
+
+Data is stored under `~/.transcribe-studio/` (override with `TRANSCRIBE_STUDIO_DATA`).
+
+## Development
+
+```bash
+uv pip install -e ".[dev]"
+uv run pytest
+```
+
+## Publishing
+
+### PyPI via uv (recommended)
+
+```bash
+uv build
+uv publish   # uses UV_PUBLISH_TOKEN or prompts for PyPI credentials
+```
+
+### PyPI via pip/twine
+
+```bash
+pip install build twine
+python -m build
+twine upload dist/*
+```
+
+### GitHub release
+
+```bash
+git tag v0.2.0
+git push origin v0.2.0
+gh release create v0.2.0 dist/*
+```
+
+## Configuration
+
+Evaluation and transcript import settings ship inside the package:
+
+- `app/config/evaluation.toml`
+- `app/config/transcript_formats.toml`
+- `app/config/languages/en.toml`
+
+## License
+
+MIT — see [LICENSE](LICENSE).
+
+## Contributing
+
+Issues and PRs welcome at [github.com/Mishkat-Quantum-Labs/transcribe-studio](https://github.com/Mishkat-Quantum-Labs/transcribe-studio).

transcribe_studio-0.2.0/app/analytics.py

@@ -0,0 +1,131 @@
+"""Dashboard and recording-level transcription analytics."""
+from __future__ import annotations
+
+import re
+from typing import Any
+
+
+def _word_count(text: str) -> int:
+    return len(re.findall(r"\S+", text or ""))
+
+
+def _segment_duration(seg: dict) -> int:
+    return max(0, seg["end_ms"] - seg["start_ms"])
+
+
+def analyze_segments(segments: list[dict], duration_ms: int | None) -> dict[str, Any]:
+    total = len(segments)
+    transcribed = sum(1 for s in segments if (s.get("transcript") or "").strip())
+    labeled = sum(1 for s in segments if (s.get("speaker") or "").strip())
+    seg_ms = sum(_segment_duration(s) for s in segments)
+    words = sum(_word_count(s.get("transcript") or "") for s in segments)
+
+    speaker_stats: dict[str, dict[str, int]] = {}
+    for s in segments:
+        name = (s.get("speaker") or "").strip() or "Unlabeled"
+        if name not in speaker_stats:
+            speaker_stats[name] = {"segments": 0, "words": 0, "duration_ms": 0}
+        speaker_stats[name]["segments"] += 1
+        speaker_stats[name]["words"] += _word_count(s.get("transcript") or "")
+        speaker_stats[name]["duration_ms"] += _segment_duration(s)
+
+    speakers = [
+        {"name": k, **v}
+        for k, v in sorted(speaker_stats.items(), key=lambda x: -x[1]["duration_ms"])
+    ]
+
+    dur = duration_ms or 0
+    coverage_pct = min(100, round(seg_ms / dur * 100)) if dur else 0
+    transcript_pct = round(transcribed / total * 100) if total else 0
+    speaker_pct = round(labeled / total * 100) if total else 0
+    avg_seg_ms = round(seg_ms / total) if total else 0
+
+    return {
+        "segment_count": total,
+        "transcribed_segments": transcribed,
+        "speaker_labeled_segments": labeled,
+        "empty_segments": total - transcribed,
+        "total_words": words,
+        "segmented_duration_ms": seg_ms,
+        "coverage_pct": coverage_pct,
+        "transcript_pct": transcript_pct,
+        "speaker_label_pct": speaker_pct,
+        "avg_segment_ms": avg_seg_ms,
+        "speakers": speakers,
+    }
+
+
+def analyze_recording(rec: dict, segments: list[dict]) -> dict[str, Any]:
+    stats = analyze_segments(segments, rec.get("duration_ms"))
+    return {
+        "id": rec["id"],
+        "title": rec["title"],
+        "duration_ms": rec.get("duration_ms"),
+        "created_at": rec.get("created_at", "")[:10],
+        "notes": rec.get("notes") or "",
+        **stats,
+    }
+
+
+def dashboard_stats(conn) -> dict[str, Any]:
+    recordings = conn.execute(
+        "SELECT id, title, duration_ms, created_at FROM recordings ORDER BY id DESC"
+    ).fetchall()
+
+    total_segments = conn.execute("SELECT COUNT(*) FROM segments").fetchone()[0]
+    total_duration = conn.execute(
+        "SELECT COALESCE(SUM(duration_ms), 0) FROM recordings"
+    ).fetchone()[0]
+
+    all_segments = conn.execute(
+        "SELECT recording_id, start_ms, end_ms, speaker, transcript FROM segments"
+    ).fetchall()
+    seg_list = [dict(s) for s in all_segments]
+    transcribed = sum(1 for s in seg_list if (s.get("transcript") or "").strip())
+    words = sum(_word_count(s.get("transcript") or "") for s in seg_list)
+    segmented_ms = sum(_segment_duration(s) for s in seg_list)
+
+    speakers = {
+        (s.get("speaker") or "").strip() or "Unlabeled"
+        for s in seg_list
+        if (s.get("transcript") or "").strip() or (s.get("speaker") or "").strip()
+    }
+
+    recording_stats = []
+    for rec in recordings:
+        rec_segs = [s for s in seg_list if s["recording_id"] == rec["id"]]
+        recording_stats.append(analyze_recording(dict(rec), rec_segs))
+
+    overall_transcript_pct = (
+        round(transcribed / total_segments * 100) if total_segments else 0
+    )
+    overall_coverage_pct = (
+        min(100, round(segmented_ms / total_duration * 100)) if total_duration else 0
+    )
+
+    return {
+        "recording_count": len(recordings),
+        "segment_count": total_segments,
+        "total_duration_ms": total_duration,
+        "segmented_duration_ms": segmented_ms,
+        "transcribed_segments": transcribed,
+        "total_words": words,
+        "unique_speakers": len(speakers),
+        "transcript_pct": overall_transcript_pct,
+        "coverage_pct": overall_coverage_pct,
+        "recordings": recording_stats,
+    }
+
+
+def fmt_duration(ms: int | None) -> str:
+    if not ms:
+        return "—"
+    s = ms / 1000
+    h = int(s // 3600)
+    m = int((s % 3600) // 60)
+    sec = s % 60
+    if h:
+        return f"{h}h {m}m"
+    if m:
+        return f"{m}m {sec:.0f}s"
+    return f"{sec:.1f}s"

transcribe_studio-0.2.0/app/config/evaluation.toml

@@ -0,0 +1,45 @@
+# Transcribe Studio - Evaluation Configuration
+# https://github.com/Mishkat-Quantum-Labs/transcribe-studio
+
+[evaluation]
+version = "1.0"
+default_language = "en"
+
+# Metrics to compute
+# Set enabled = false to skip a metric
+# weight is used for weighted averaging in overall score
+
+[evaluation.metrics]
+
+[evaluation.metrics.wer]
+enabled = true
+weight = 1.0
+case_sensitive = false
+description = "Word Error Rate - standard ASR metric"
+
+[evaluation.metrics.cer]
+enabled = false
+weight = 0.0
+description = "Character Error Rate - useful for character-level languages"
+
+[evaluation.metrics.semantic_score]
+enabled = true
+weight = 0.5
+description = "Semantic equivalence score - partial credit for meaning"
+
+# Text normalization settings
+# These apply before metric calculation
+
+[evaluation.normalization]
+lowercase = true
+trim_whitespace = true
+remove_punctuation = false
+normalize_quotes = true
+remove_special_chars = false
+
+# UI Settings
+
+[evaluation.ui]
+show_detailed_breakdown = true
+highlight_errors = true
+color_scheme = "auto"

transcribe_studio-0.2.0/app/config/languages/en.toml

@@ -0,0 +1,242 @@
+# English Language Configuration
+# Semantic equivalence rules for English
+
+[language]
+code = "en"
+name = "English"
+normalizer_class = "en"
+
+# ============================================================
+# SEMANTIC MATCHING RULES
+# ============================================================
+# These rules define phrases that are semantically equivalent
+# even when they differ in exact wording.
+#
+# Each rule has:
+# - variants: list of alternative phrasings
+# - canonical: the "standard" form to compare against
+# - weight: 0.0-1.0, confidence of equivalence
+#
+# Matching works bidirectionally:
+# "gonna" matches "going to" and vice versa
+# ============================================================
+
+[[semantic_matchers.group]]
+name = "contractions_informal"
+description = "Contractions and informal speech → formal forms"
+enabled = true
+
+[[semantic_matchers.group.rule]]
+variants = ["gonna", "gon na", "gunna", "gonna"]
+canonical = "going to"
+weight = 0.95
+
+[[semantic_matchers.group.rule]]
+variants = ["wanna", "wanner"]
+canonical = "want to"
+weight = 0.95
+
+[[semantic_matchers.group.rule]]
+variants = ["gotta", "got a"]
+canonical = "got to"
+weight = 0.9
+
+[[semantic_matchers.group.rule]]
+variants = ["kinda", "kind of"]
+canonical = "kind of"
+weight = 0.9
+
+[[semantic_matchers.group.rule]]
+variants = ["sorta", "sort of"]
+canonical = "sort of"
+weight = 0.9
+
+[[semantic_matchers.group.rule]]
+variants = ["outta", "out of"]
+canonical = "out of"
+weight = 0.95
+
+[[semantic_matchers.group.rule]]
+variants = ["lemme", "let me"]
+canonical = "let me"
+weight = 0.95
+
+[[semantic_matchers.group.rule]]
+variants = ["gimme", "give me"]
+canonical = "give me"
+weight = 0.95
+
+[[semantic_matchers.group.rule]]
+variants = ["dunno", "dont know", "do not know", "don't know"]
+canonical = "do not know"
+weight = 0.9
+
+[[semantic_matchers.group.rule]]
+variants = ["coulda", "could have", "could've"]
+canonical = "could have"
+weight = 0.95
+
+[[semantic_matchers.group.rule]]
+variants = ["woulda", "would have", "would've"]
+canonical = "would have"
+weight = 0.95
+
+[[semantic_matchers.group.rule]]
+variants = ["shoulda", "should have", "should've"]
+canonical = "should have"
+weight = 0.95
+
+[[semantic_matchers.group.rule]]
+variants = ["lotsa", "lots of"]
+canonical = "lots of"
+weight = 0.95
+
+[[semantic_matchers.group.rule]]
+variants = ["cause", "cos", "cuz"]
+canonical = "because"
+weight = 0.85
+
+[[semantic_matchers.group.rule]]
+variants = ["nvm", "nvr", "nevermind", "never mind"]
+canonical = "never mind"
+weight = 0.95
+
+[[semantic_matchers.group.rule]]
+variants = ["thru", "through"]
+canonical = "through"
+weight = 0.98
+
+[[semantic_matchers.group.rule]]
+variants = ["u", "you"]
+canonical = "you"
+weight = 0.8
+
+[[semantic_matchers.group.rule]]
+variants = ["ur", "you're", "your"]
+canonical = "your"
+weight = 0.7
+
+[[semantic_matchers.group.rule]]
+variants = ["ok", "okay", "ok"]
+canonical = "okay"
+weight = 1.0
+
+[[semantic_matchers.group.rule]]
+variants = ["yeah", "yes", "yea", "yah"]
+canonical = "yes"
+weight = 0.95
+
+[[semantic_matchers.group.rule]]
+variants = ["nope", "no", "nah"]
+canonical = "no"
+weight = 0.95
+
+[[semantic_matchers.group.rule]]
+variants = ["alright", "all right", "allright"]
+canonical = "all right"
+weight = 1.0
+
+[[semantic_matchers.group.rule]]
+variants = ["gonna", "goin", "goin to", "going"]
+canonical = "going"
+weight = 0.8
+
+[[semantic_matchers.group]]
+name = "repeated_sounds"
+description = "Stuttered/repeated sounds - common in spontaneous speech"
+enabled = true
+
+[[semantic_matchers.group.rule]]
+variants = ["um", "uh", "er", "erm"]
+canonical = ""
+weight = 0.5
+
+[[semantic_matchers.group]]
+name = "common_confusions"
+description = "Common ASR/LLM transcription confusions"
+enabled = true
+
+[[semantic_matchers.group.rule]]
+variants = ["i am", "i'm", "im"]
+canonical = "i am"
+weight = 1.0
+
+[[semantic_matchers.group.rule]]
+variants = ["you know", "yknow", "y'know"]
+canonical = "you know"
+weight = 0.9
+
+[[semantic_matchers.group.rule]]
+variants = ["like", "like like"]
+canonical = "like"
+weight = 0.7
+
+[[semantic_matchers.group]]
+name = "numbers"
+description = "Number word ↔ digit equivalence"
+enabled = true
+
+[[semantic_matchers.group.rule]]
+variants = ["for", "four"]
+canonical = "four"
+weight = 0.9
+
+[[semantic_matchers.group.rule]]
+variants = ["to", "two", "too"]
+canonical = "two"
+weight = 0.8
+
+[[semantic_matchers.group]]
+name = "contractions"
+description = "Standard English contractions"
+enabled = true
+
+[[semantic_matchers.group.rule]]
+variants = ["don't", "do not"]
+canonical = "do not"
+weight = 1.0
+
+[[semantic_matchers.group.rule]]
+variants = ["can't", "cannot"]
+canonical = "cannot"
+weight = 1.0
+
+[[semantic_matchers.group.rule]]
+variants = ["won't", "will not"]
+canonical = "will not"
+weight = 1.0
+
+[[semantic_matchers.group.rule]]
+variants = ["i've", "i have"]
+canonical = "i have"
+weight = 1.0
+
+[[semantic_matchers.group.rule]]
+variants = ["i'll", "i will"]
+canonical = "i will"
+weight = 1.0
+
+[[semantic_matchers.group.rule]]
+variants = ["it's", "it is", "its"]
+canonical = "it is"
+weight = 0.95
+
+[[semantic_matchers.group.rule]]
+variants = ["that's", "that is"]
+canonical = "that is"
+weight = 1.0
+
+[[semantic_matchers.group.rule]]
+variants = ["there's", "there is"]
+canonical = "there is"
+weight = 1.0
+
+[[semantic_matchers.group.rule]]
+variants = ["here's", "here is"]
+canonical = "here is"
+weight = 1.0
+
+[[semantic_matchers.group.rule]]
+variants = ["what's", "what is"]
+canonical = "what is"
+weight = 1.0