interview-transcription

Name: interview-transcription
Author: jamditis

136installs57forks

How to install interview-transcription?

jamditis/claude-skills-journalism/interview-transcription

$npx -y skills add jamditis/claude-skills-journalism --skill interview-transcription

Installs into the current project.

Summary

Transcription workflows, recording management, and quote extraction for journalists. Use when processing audio/video recordings, generating transcripts with timestamps, extracting quotes for fact-checking, or building source-and-recording databases. For interview question design

Files · 1

View on GitHub

SKILL.md

1	# Interview transcription and management
2
3	Practical workflows for journalists managing interviews from preparation through publication.
4
5	## When to activate
6
7	- Preparing questions for an interview
8	- Processing audio/video recordings
9	- Creating or managing transcripts
10	- Organizing notes from multiple sources
11	- Building a source relationship database
12	- Generating timestamped quotes for fact-checking
13	- Converting recordings to publishable quotes
14
15	## Recording setup for transcription
16
17	For pre-interview research, question design, attribution agreements, and consent scripts, use the interview-prep skill. The notes here cover only the recording configuration that affects transcription quality.
18
19	```python
20	# Standard recording configuration for clean transcription
21	RECORDING_SETTINGS = {
22	'format': 'wav', # Lossless for transcription
23	'sample_rate': 16000, # Whisper resamples to 16k anyway; 16k saves disk
24	'channels': 1, # Mono is fine for speech; stereo only if mics are positionally distinct
25	'backup': True, # Always run a backup recorder
26	}
27
28	# File naming convention
29	# YYYY-MM-DD_source-lastname_topic.wav
30	# Example: 2026-05-08_smith_budget-hearing.wav
31	```
32
33	Two-device rule. Always record on two devices. Phone as backup minimum. If using a wireless lav mic, the recorder built into the lav unit is one device; the phone running a backup app is the second.
34
35	Mono is preferred unless each speaker has their own dedicated microphone routed to a distinct channel. Stereo with both speakers bleeding into both channels is worse for diarization than clean mono.
36
37	## Transcription workflows
38
39	### Automated transcription pipeline
40
41	Vanilla OpenAI Whisper transcribes audio to text but does not assign speaker labels. To get diarized output ("Speaker 1:" / "Speaker 2:" / etc.) you need a tool that combines Whisper with a diarization model — typically WhisperX (`m-bain/whisperX`), which wraps faster-whisper transcription with pyannote.audio diarization and produces word-level timestamps with speaker IDs in one pass.
42
43	```python
44	from pathlib import Path
45	import subprocess
46	import json
47
48	def transcribe_interview(
49	audio_path: str,
50	output_dir: str = "./transcripts",
51	diarize: bool = True,
52	hf_token: str \| None = None,
53	min_speakers: int = 2,
54	max_speakers: int = 2,
55	) -> dict:
56	"""
57	Transcribe an interview using WhisperX (Whisper + pyannote diarization).
58	Returns a transcript with word-level timestamps and speaker labels.
59
60	Diarization needs a Hugging Face token with access to the pyannote
61	speaker-diarization-3.1 model. Accept the model EULA at
62	huggingface.co/pyannote/speaker-diarization-3.1 once, then pass the token.
63	"""
64	Path(output_dir).mkdir(exist_ok=True)
65
66	cmd = [
67	'whisperx', audio_path,
68	'--model', 'large-v3',
69	'--output_format', 'json',
70	'--output_dir', output_dir,
71	'--language', 'en',
72	'--compute_type', 'int8', # CPU-friendly; use 'float16' on GPU
73	'--min_speakers', str(min_speakers),
74	'--max_speakers', str(max_speakers),
75	]
76
77	if diarize:
78	cmd.append('--diarize')
79	if hf_token:
80	cmd += ['--hf_token', hf_token]
81
82	subprocess.run(cmd, check=True, capture_output=True)
83
84	json_path = Path(output_dir) / f"{Path(audio_path).stem}.json"
85	with open(json_path) as f:
86	return json.load(f)
87
88	def format_for_editing(transcript: dict) -> str:
89	"""Convert to journalist-friendly format with timestamps."""
90	lines = []
91	for segment in transcript.get('segments', []):
92	timestamp = format_timestamp(segment['start'])
93	text = segment['text'].strip()
94	lines.append(f"[{timestamp}] {text}")
95	return '\n\n'.join(lines)
96
97	def format_timestamp(seconds: float) -> str:
98	"""Convert seconds to HH:MM:SS format."""
99	h = int(seconds // 3600)
100	m = int((seconds % 3600) // 60)
101	s = int(seconds % 60)
102	return f"{h:02d}:{m:02d}:{s:02d}"
103	```
104
105	Falling back to plain Whisper. If diarization is overkill or you can't get a Hugging Face token, drop the `--diarize` flag — the model still produces accurate timestamped transcription and you label speakers manually based on context. `faster-whisper` (CTranslate2 backend) is the speed-optimized variant and works the same way at the CLI. `whisper.cpp` is the C++ port for resource-constrained machines (Raspberry Pi, older laptops); it doesn't include diarization but runs the small/medium models on CPU comfortably.
106
107	### Manual transcription template
108
109	For sensitive interviews or when AI transc

People who liked this also viewed

6 picks

Type

interview-transcription

136installs57forks

View on GitHub

How to install interview-transcription?

jamditis/claude-skills-journalism/interview-transcription

$npx -y skills add jamditis/claude-skills-journalism --skill interview-transcription

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Interview transcription and management
2
3	Practical workflows for journalists managing interviews from preparation through publication.
4
5	## When to activate
6
7	- Preparing questions for an interview
8	- Processing audio/video recordings
9	- Creating or managing transcripts
10	- Organizing notes from multiple sources
11	- Building a source relationship database
12	- Generating timestamped quotes for fact-checking
13	- Converting recordings to publishable quotes
14
15	## Recording setup for transcription
16
17	For pre-interview research, question design, attribution agreements, and consent scripts, use the interview-prep skill. The notes here cover only the recording configuration that affects transcription quality.
18
19	```python
20	# Standard recording configuration for clean transcription
21	RECORDING_SETTINGS = {
22	'format': 'wav', # Lossless for transcription
23	'sample_rate': 16000, # Whisper resamples to 16k anyway; 16k saves disk
24	'channels': 1, # Mono is fine for speech; stereo only if mics are positionally distinct
25	'backup': True, # Always run a backup recorder
26	}
27
28	# File naming convention
29	# YYYY-MM-DD_source-lastname_topic.wav
30	# Example: 2026-05-08_smith_budget-hearing.wav
31	```
32
33	Two-device rule. Always record on two devices. Phone as backup minimum. If using a wireless lav mic, the recorder built into the lav unit is one device; the phone running a backup app is the second.
34
35	Mono is preferred unless each speaker has their own dedicated microphone routed to a distinct channel. Stereo with both speakers bleeding into both channels is worse for diarization than clean mono.
36
37	## Transcription workflows
38
39	### Automated transcription pipeline
40
41	Vanilla OpenAI Whisper transcribes audio to text but does not assign speaker labels. To get diarized output ("Speaker 1:" / "Speaker 2:" / etc.) you need a tool that combines Whisper with a diarization model — typically WhisperX (`m-bain/whisperX`), which wraps faster-whisper transcription with pyannote.audio diarization and produces word-level timestamps with speaker IDs in one pass.
42
43	```python
44	from pathlib import Path
45	import subprocess
46	import json
47
48	def transcribe_interview(
49	audio_path: str,
50	output_dir: str = "./transcripts",
51	diarize: bool = True,
52	hf_token: str \| None = None,
53	min_speakers: int = 2,
54	max_speakers: int = 2,
55	) -> dict:
56	"""
57	Transcribe an interview using WhisperX (Whisper + pyannote diarization).
58	Returns a transcript with word-level timestamps and speaker labels.
59
60	Diarization needs a Hugging Face token with access to the pyannote
61	speaker-diarization-3.1 model. Accept the model EULA at
62	huggingface.co/pyannote/speaker-diarization-3.1 once, then pass the token.
63	"""
64	Path(output_dir).mkdir(exist_ok=True)
65
66	cmd = [
67	'whisperx', audio_path,
68	'--model', 'large-v3',
69	'--output_format', 'json',
70	'--output_dir', output_dir,
71	'--language', 'en',
72	'--compute_type', 'int8', # CPU-friendly; use 'float16' on GPU
73	'--min_speakers', str(min_speakers),
74	'--max_speakers', str(max_speakers),
75	]
76
77	if diarize:
78	cmd.append('--diarize')
79	if hf_token:
80	cmd += ['--hf_token', hf_token]
81
82	subprocess.run(cmd, check=True, capture_output=True)
83
84	json_path = Path(output_dir) / f"{Path(audio_path).stem}.json"
85	with open(json_path) as f:
86	return json.load(f)
87
88	def format_for_editing(transcript: dict) -> str:
89	"""Convert to journalist-friendly format with timestamps."""
90	lines = []
91	for segment in transcript.get('segments', []):
92	timestamp = format_timestamp(segment['start'])
93	text = segment['text'].strip()
94	lines.append(f"[{timestamp}] {text}")
95	return '\n\n'.join(lines)
96
97	def format_timestamp(seconds: float) -> str:
98	"""Convert seconds to HH:MM:SS format."""
99	h = int(seconds // 3600)
100	m = int((seconds % 3600) // 60)
101	s = int(seconds % 60)
102	return f"{h:02d}:{m:02d}:{s:02d}"
103	```
104
105	Falling back to plain Whisper. If diarization is overkill or you can't get a Hugging Face token, drop the `--diarize` flag — the model still produces accurate timestamped transcription and you label speakers manually based on context. `faster-whisper` (CTranslate2 backend) is the speed-optimized variant and works the same way at the CLI. `whisper.cpp` is the C++ port for resource-constrained machines (Raspberry Pi, older laptops); it doesn't include diarization but runs the small/medium models on CPU comfortably.
106
107	### Manual transcription template
108
109	For sensitive interviews or when AI transc

People who liked this also viewed

6 picks

Type