Luel-ai/luel-conversational-speech-samples

Conversational Speech Samples (Luel)

License: All Rights Reserved. Proprietary. Access only for authorized parties; no redistribution or use without permission. See LICENSE.

A multilingual conversational speech dataset containing 52 dialogue sessions across 4 languages: German, English, French, and Japanese. Each session includes: combined audio (both speakers in one file), speaker-separated audio (one WAV per speaker), and a JSON transcript with full text, speaker-diarized utterances, word-level timestamps, and metadata.

---

Quick Stats

Metric	Value
Languages	4 (German, English, French, Japanese)
Total sessions	52
Total duration	~29h 52m
Total utterances	16,059
Total words (with timestamps)	275,048
Total audio files	156 WAV
Total JSON files	52
Total size on disk	28.9 GB
Speakers per session	2
Sample rate	48 kHz

---

Per-Language Breakdown

Language	Code	Sessions	Duration	Utterances	Words	Audio Files	Size
German	de	13	~6h 31m	1,571	64,752	39	6.3 GB
English	en	13	~5h 07m	3,304	41,462	39	5.0 GB
French	fr	13	~5h 43m	1,213	39,633	39	5.5 GB
Japanese	ja	13	~12h 31m	9,971	129,201	39	12.1 GB
Total		52	~29h 52m	16,059	275,048	156	28.9 GB

---

Directory Structure

Conversational_Speech_Samples_Luel/
  README.md
  manifest.json
  de/
    arctic_cruises/
      arctic_cruises.json
      arctic_cruises.wav              (combined audio, both speakers)
      arctic_cruises_speaker1.wav     (speaker 1 only)
      arctic_cruises_speaker2.wav     (speaker 2 only)
    business_model_canvas/
      ...
  en/
    session_01/
      session_01.json
      session_01.wav
      ...
  fr/
    boutique_shopping/
      ...
  ja/
    うん/
      ...

Each session folder contains exactly 4 files: one JSON transcript, one combined audio file, and two speaker-separated audio files.

File	Description
{name}.json	Transcript (metadata, speaker-diarized text, word-level timestamps)
{name}.wav	Combined audio (both speakers)
{name}_speaker1.wav	Speaker-separated audio: Speaker 1 only
{name}_speaker2.wav	Speaker-separated audio: Speaker 2 only

---

JSON Schema

Every JSON file has the following top-level structure:

{
  "UUID": "9eecbeed-54f9-4b62-ac36-f5884fbdd883",
  "audioURL": "en/session_01/session_01.wav",
  "name": "session_01",
  "durationSeconds": 3612.416,
  "language": "English",
  "category": "spontaneous_discussions",
  "topic": "upgrading microphone and audio equipment setup",
  "numSpeakers": 2,
  "metadata": {
    "language": "English",
    "country": "USA",
    "region": "New York",
    "environment": "quiet",
    "device": "PC",
    "topic": "upgrading microphone and audio equipment setup"
  },
  "transcript": "Full transcript as a single string...",
  "speakerTranscript": [
    {
      "speaker": "Speaker1",
      "text": "You know what happened now?",
      "startTime": 8.72,
      "endTime": 9.76,
      "words": [
        { "text": "You", "startTime": 8.72, "endTime": 8.84 },
        { "text": "know", "startTime": 8.84, "endTime": 9.0 },
        { "text": "what", "startTime": 9.0, "endTime": 9.16 },
        { "text": "happened", "startTime": 9.16, "endTime": 9.48 },
        { "text": "now?", "startTime": 9.48, "endTime": 9.76 }
      ]
    }
  ],
  "sampleRate": 48000
}

Top-Level Fields

Field	Type	Description
UUID	string	Unique identifier (UUID v4)
audioURL	string	Relative path to combined audio: {lang}/{name}/{name}.wav
name	string	Session name (matches folder name)
durationSeconds	number	Total duration of the combined audio in seconds
language	string	Language name: German, English, French, or Japanese
category	string	One of: formal_meetings, spontaneous_discussions
topic	string	Short description of the conversation topic
numSpeakers	integer	Number of unique speakers (always 2)
metadata	object	Recording metadata (see below)
transcript	string	Full conversation transcript as a single string
speakerTranscript	array	Speaker-diarized utterances with word-level timestamps
sampleRate	integer	Audio sample rate in Hz (48000)

Metadata Fields

Field	Type	Description
language	string	Recording language
country	string	Country of origin of the speakers
region	string	Region / city
environment	string	Recording environment (e.g., quiet)
device	string	Recording device (PC, Laptop, Phone)
topic	string	Conversation topic

Utterance Fields (speakerTranscript[*])

Field	Type	Description
speaker	string	Speaker identifier (Speaker1, Speaker2)
text	string	Utterance text
startTime	number	Start time in seconds from audio start
endTime	number	End time in seconds from audio start
words	array	Word-level timestamps (see below)

Word Fields (speakerTranscript[*].words[*])

Field	Type	Description
text	string	Single word or token
startTime	number	Word start time in seconds
endTime	number	Word end time in seconds

---

Categories

Category	Sessions	Description
formal_meetings	20	Structured business discussions, negotiations, presentations
spontaneous_discussions	32	Casual, unscripted conversations on everyday topics

---

Session Listing

German (de) -- 13 sessions, 6h 31m

Session	Topic	Category	Duration
arctic_cruises	arctic cruises	spontaneous_discussions	25:21
business_model_canvas	Canvas Business Model	formal_meetings	30:06
business_optimization	How to optimize a Business	formal_meetings	30:20
business_structure	Structure of a Business	formal_meetings	30:22
customer_service	Customer Service	formal_meetings	25:33
expat_life	Friends catching up, expat life	spontaneous_discussions	29:59
fashion_sustainability	Fashion and sustainability	spontaneous_discussions	21:32
food	food	spontaneous_discussions	25:27
getting_to_know	Getting to know each other, life and travel	spontaneous_discussions	57:21
house_purchase	house purchase	formal_meetings	25:22
latin_america	Latin america	spontaneous_discussions	26:16
pos_planning	POS Planung	formal_meetings	38:11
working_abroad	maternity leave and work conditions usa vs germany	spontaneous_discussions	25:00

English (en) -- 13 sessions, 5h 07m

Session	Topic	Category	Duration
session_01	upgrading microphone and audio equipment setup	spontaneous_discussions	1:00:12
session_02	school, scheduling, and study plans	spontaneous_discussions	15:17
session_03	pets, life, and casual conversation	spontaneous_discussions	30:01
session_04	playing cities: skylines	spontaneous_discussions	16:18
session_05	online gaming and scams	spontaneous_discussions	29:25
session_06	playing a video game	spontaneous_discussions	44:53
session_07	discussing games and piracy	spontaneous_discussions	28:44
session_08	playing and learning colonist	spontaneous_discussions	32:55
session_09	discussing soccer video game tactics	spontaneous_discussions	26:22
session_10	feeling overwhelmed and stressed	spontaneous_discussions	5:03
session_11	hotel cancellation and customer support	formal_meetings	7:38
session_12	catching up with friends	spontaneous_discussions	5:23
session_13	catching up with friends	spontaneous_discussions	5:14

French (fr) -- 13 sessions, 5h 43m

Session	Topic	Category	Duration
boutique_shopping	Cliente et vendeuse dans une boutique	formal_meetings	30:06
budgeting	Budgeting	formal_meetings	30:23
call_center	Call center Work call	formal_meetings	30:10
car_sales	CARS SALE	formal_meetings	23:24
clothing_store	Magazin de vêtements	formal_meetings	30:12
cultural_differences	Cultural differences	spontaneous_discussions	30:33
daily_life	Daily life and stress	spontaneous_discussions	20:45
food_memories	Food and memories	spontaneous_discussions	20:59
freelancers_routine	Freelancers' routine	spontaneous_discussions	20:56
house_purchase	house purchase	formal_meetings	24:32
japan_tourism	Tourism: Japan and China	spontaneous_discussions	29:24
phones_impact	Phones' impact	spontaneous_discussions	20:54
restaurant_service	Présentation des plats et recommandations du Chef	formal_meetings	30:36

Japanese (ja) -- 13 sessions, 12h 31m

Session	Topic	Category	Duration
いやねー	買い物、健康、旅行、宗教	spontaneous_discussions	1:00:45
うん	旅行中の出会いと人物評	spontaneous_discussions	59:32
うんいや_1	タワマンでの子育てと人間関係	spontaneous_discussions	59:26
うんうーん	国井さんの人物像と音楽の好み	formal_meetings	1:00:17
うんお疲れ様です	アイドル引退と結婚の憶測	spontaneous_discussions	1:00:02
うんそう	人生の目標とキャリア選択	spontaneous_discussions	59:47
うんはい	サッカーチームの戦術と選手	formal_meetings	1:00:20
おはようございます	日常生活、仕事、性別の役割、人間関係	spontaneous_discussions	30:08
こんど	聴覚過敏、電車、宝塚劇場	spontaneous_discussions	1:00:11
はいあの	脚本スクールの体験談	formal_meetings	1:00:12
はいうん	親子関係と後悔	formal_meetings	1:00:26
はいそうですね_1	年金と資産運用	formal_meetings	1:00:15
はいどうも	自己紹介と職務経歴	formal_meetings	59:23

---

Audio Format

All 156 WAV files share the same format:

Property	Value
Sample rate	48,000 Hz
Bit depth	16-bit PCM
Channels	1 (mono)
Format	WAV (RIFF)

Each session folder contains three audio files:

• Combined audio ({name}.wav): mono mix of both speakers
• Speaker 1 ({name}_speaker1.wav): isolated audio for Speaker 1
• Speaker 2 ({name}_speaker2.wav): isolated audio for Speaker 2

---

Notes

• All timestamps are in seconds relative to the start of the combined audio file.
• The audioURL field is a relative path in the format {lang}/{name}/{name}.wav.
• All sessions have exactly 2 speakers.