Card: add DDS source (194,400 rows / ~252h), CC-BY-NC-4.0, real clean-studio enhancement GT
Browse files
README.md
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
---
|
| 2 |
-
license: cc-by-
|
| 3 |
language:
|
| 4 |
- en
|
| 5 |
- zh
|
|
@@ -31,31 +31,36 @@ pairs four short **24 kHz mono FLAC** clips with aligned transcripts:
|
|
| 31 |
|
| 32 |
- an **environment** sample (different speaker, same acoustic scene),
|
| 33 |
- a **speaker** reference (same speaker as the target utterance),
|
| 34 |
-
- a **speaker-enhanced** copy of the reference (MossFormer2
|
|
|
|
| 35 |
- the target **speech** to synthesise,
|
| 36 |
|
| 37 |
so a model can learn to generate an utterance with both a specified voice and a
|
| 38 |
specified environment. This release supersedes the earlier
|
| 39 |
[Env-TTS-SD-Corpus](https://huggingface.co/datasets/humanify/Env-TTS-SD-Corpus)
|
| 40 |
with a richer schema, higher sample rate, per-clip ASR for all three contexts,
|
| 41 |
-
|
|
|
|
| 42 |
|
| 43 |
## Dataset statistics
|
| 44 |
|
| 45 |
| metric | value |
|
| 46 |
| --- | ---: |
|
| 47 |
-
| **rows** | 189,918 |
|
| 48 |
-
| **on-disk size** | ~
|
| 49 |
-
| **Σ `speech_duration`** | **
|
| 50 |
-
| **Σ (`environment_audio_duration` + `speaker_audio_duration` + `speech_duration`)** | **
|
| 51 |
|
| 52 |
-
|
| 53 |
-
|
|
|
|
|
|
|
| 54 |
|
| 55 |
### Rows by source dataset
|
| 56 |
|
| 57 |
| `dataset` | rows |
|
| 58 |
| --- | ---: |
|
|
|
|
| 59 |
| `m3sd` | 108,708 |
|
| 60 |
| `aishell4` | 27,924 |
|
| 61 |
| `alimeeting` | 25,087 |
|
|
@@ -67,29 +72,30 @@ columns across all 239 parquet shards (no audio decode).
|
|
| 67 |
|
| 68 |
| column | type | description |
|
| 69 |
| --- | --- | --- |
|
| 70 |
-
| `environment_audio_source` | binary (FLAC 24 kHz mono) | acoustic-scene reference, 2.5–15 s, from a **different speaker** in the same
|
| 71 |
| `environment_audio_duration` | float32 | seconds |
|
| 72 |
-
| `environment_audio_text` | string | transcript of the environment clip (gold
|
| 73 |
| `speaker_audio_source` | binary (FLAC 24 kHz mono) | speaker-identity reference, 2.5–15 s, **same speaker** as `speech` |
|
| 74 |
| `speaker_audio_duration` | float32 | seconds |
|
| 75 |
| `speaker_audio_text` | string | transcript of the speaker reference clip |
|
| 76 |
-
| `speaker_audio_source_enhanced` | binary (FLAC 24 kHz mono) | MossFormer2-enhanced
|
| 77 |
| `text` | string | transcript of `speech` |
|
| 78 |
| `speech` | binary (FLAC 24 kHz mono) | target utterance, 3–15 s |
|
| 79 |
| `speech_duration` | float32 | seconds |
|
| 80 |
| `language` | string | `zh` / `en` / `auto` |
|
| 81 |
-
| `dataset` | string | `m3sd` / `aishell4` / `msdwild` / `chime6` / `ami` / `alimeeting` |
|
| 82 |
-
| `conversation_id` | string | unique within the source dataset |
|
| 83 |
-
| `speaker_id` | string | within-
|
| 84 |
-
| `env_id` | string | acoustic-scene identifier (
|
| 85 |
| `text_source` | string | `original`, `asr`, or `mixed` |
|
| 86 |
-
| `asr_token_count` | int32 | Qwen3-ASR token count for `speech` |
|
| 87 |
| `asr_mean_logprob` | float32 | mean log-prob per token for `speech` |
|
| 88 |
|
| 89 |
## Source corpora
|
| 90 |
|
| 91 |
-
| dataset |
|
| 92 |
| --- | ---: | --- | --- |
|
|
|
|
| 93 |
| **M3SD** (Wu et al., 2025) | 770 | zh / en mixed | ❌ → Qwen3-ASR |
|
| 94 |
| **AISHELL-4** (Fu et al., 2021) | 120 | zh | ✅ TextGrid |
|
| 95 |
| **MSDWILD** (Liu et al., 2022) | 80 | zh / en mixed | ❌ → Qwen3-ASR |
|
|
@@ -97,40 +103,48 @@ columns across all 239 parquet shards (no audio decode).
|
|
| 97 |
| **AMI** (SDM, diarizers-community) | ~100 | en | ❌ → Qwen3-ASR |
|
| 98 |
| **AliMeeting** (OpenSLR 119, far ch.0) | ~120 | zh | ✅ TextGrid |
|
| 99 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 100 |
## Processing pipeline
|
| 101 |
|
| 102 |
Built with the streaming pipeline in
|
| 103 |
-
[`env-tts-data-pipeline`](https://github.com/ChristianYang/env-tts-data-pipeline)
|
| 104 |
-
|
|
|
|
| 105 |
|
| 106 |
1. **download** — stream each source conversation (HF mirrors, OpenSLR tar
|
| 107 |
-
streams,
|
| 108 |
-
sentinel when audio is ready.
|
| 109 |
2. **process** — resample to **24 kHz mono**, walk diarisation turns, emit
|
| 110 |
3–15 s `speech` slices with a same-speaker reference (≥2.5 s) and a
|
| 111 |
-
different-speaker environment slice (≥2.5 s
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
DAIEN-style RIR+noise augmentation on the speaker reference is **disabled** in
|
| 122 |
-
this release; environment/speaker decoupling relies on cross-speaker env sampling
|
| 123 |
-
plus MossFormer2 enhancement instead.
|
| 124 |
|
| 125 |
## Licensing
|
| 126 |
|
| 127 |
-
Released under **CC-BY-
|
| 128 |
-
sources. In particular:
|
| 129 |
|
|
|
|
|
|
|
| 130 |
- **M3SD** — academic / non-commercial research only.
|
| 131 |
- **MSDWILD** — X-LANCE research-only agreement.
|
| 132 |
- AISHELL-4 (Apache-2.0), CHiME-6 (CC-BY-SA-4.0), AMI, and AliMeeting carry
|
| 133 |
-
their respective open/research terms.
|
| 134 |
|
| 135 |
Redistributing extracted audio requires complying with each upstream licence.
|
| 136 |
|
|
@@ -139,6 +153,13 @@ Redistributing extracted audio requires complying with each upstream licence.
|
|
| 139 |
Please cite the source papers when using this corpus:
|
| 140 |
|
| 141 |
```bibtex
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 142 |
@article{wu2025m3sd,
|
| 143 |
title={M3SD: Multi-modal, Multi-scenario and Multi-language Speaker
|
| 144 |
Diarization Dataset},
|
|
@@ -172,7 +193,8 @@ Please cite the source papers when using this corpus:
|
|
| 172 |
```
|
| 173 |
|
| 174 |
ASR re-labelling uses [Qwen3-ASR-1.7B](https://huggingface.co/Qwen/Qwen3-ASR-1.7B).
|
| 175 |
-
Speaker enhancement uses MossFormer2 (ClearVoice).
|
|
|
|
| 176 |
|
| 177 |
## Loading
|
| 178 |
|
|
@@ -181,21 +203,20 @@ from datasets import load_dataset
|
|
| 181 |
|
| 182 |
ds = load_dataset("ChristianYang/Env-TTS-Clean", split="train", streaming=True)
|
| 183 |
row = next(iter(ds))
|
| 184 |
-
print(row["text"])
|
| 185 |
-
print(row["speech_duration"])
|
| 186 |
# Audio columns decode automatically when accessed (24 kHz mono).
|
|
|
|
|
|
|
|
|
|
| 187 |
```
|
| 188 |
|
| 189 |
## Files on disk
|
| 190 |
|
| 191 |
```
|
| 192 |
data/
|
| 193 |
-
group_00000/
|
| 194 |
manifest.json
|
| 195 |
data_000000.parquet
|
| 196 |
-
data_000001.parquet
|
| 197 |
-
...
|
| 198 |
-
group_00001/
|
| 199 |
...
|
| 200 |
```
|
| 201 |
|
|
|
|
| 1 |
---
|
| 2 |
+
license: cc-by-nc-4.0
|
| 3 |
language:
|
| 4 |
- en
|
| 5 |
- zh
|
|
|
|
| 31 |
|
| 32 |
- an **environment** sample (different speaker, same acoustic scene),
|
| 33 |
- a **speaker** reference (same speaker as the target utterance),
|
| 34 |
+
- a **speaker-enhanced** copy of the reference (MossFormer2 enhancement — or, for
|
| 35 |
+
the **DDS** source, the **real clean-studio recording** of the speaker reference),
|
| 36 |
- the target **speech** to synthesise,
|
| 37 |
|
| 38 |
so a model can learn to generate an utterance with both a specified voice and a
|
| 39 |
specified environment. This release supersedes the earlier
|
| 40 |
[Env-TTS-SD-Corpus](https://huggingface.co/datasets/humanify/Env-TTS-SD-Corpus)
|
| 41 |
with a richer schema, higher sample rate, per-clip ASR for all three contexts,
|
| 42 |
+
two meeting sources (AMI, AliMeeting), and a controlled read-speech source
|
| 43 |
+
(**DDS / DAPS**) that contributes real clean-studio enhancement ground truth.
|
| 44 |
|
| 45 |
## Dataset statistics
|
| 46 |
|
| 47 |
| metric | value |
|
| 48 |
| --- | ---: |
|
| 49 |
+
| **rows** | **384,318** (189,918 diarization + 194,400 DDS) |
|
| 50 |
+
| **on-disk size** | ~197 GB |
|
| 51 |
+
| **Σ `speech_duration`** | **~680 h** (428.6 h diarization + ~251.7 h DDS) |
|
| 52 |
+
| **Σ (`environment_audio_duration` + `speaker_audio_duration` + `speech_duration`)** | **~1,964 h** (1,252.1 h + ~712 h DDS) |
|
| 53 |
|
| 54 |
+
Diarization-source totals were computed exactly on 2026-05-27 (239 shards). The
|
| 55 |
+
**DDS** portion (194,400 rows, added 2026-06-09; train groups `group_00061`–
|
| 56 |
+
`group_00122`) is estimated from a sampled shard (mean per row: speech 4.66 s,
|
| 57 |
+
env 4.24 s, speaker 4.29 s).
|
| 58 |
|
| 59 |
### Rows by source dataset
|
| 60 |
|
| 61 |
| `dataset` | rows |
|
| 62 |
| --- | ---: |
|
| 63 |
+
| `dds` | 194,400 |
|
| 64 |
| `m3sd` | 108,708 |
|
| 65 |
| `aishell4` | 27,924 |
|
| 66 |
| `alimeeting` | 25,087 |
|
|
|
|
| 72 |
|
| 73 |
| column | type | description |
|
| 74 |
| --- | --- | --- |
|
| 75 |
+
| `environment_audio_source` | binary (FLAC 24 kHz mono) | acoustic-scene reference, 2.5–15 s, from a **different speaker** in the same scene |
|
| 76 |
| `environment_audio_duration` | float32 | seconds |
|
| 77 |
+
| `environment_audio_text` | string | transcript of the environment clip (gold / Qwen3-ASR / DAPS script) |
|
| 78 |
| `speaker_audio_source` | binary (FLAC 24 kHz mono) | speaker-identity reference, 2.5–15 s, **same speaker** as `speech` |
|
| 79 |
| `speaker_audio_duration` | float32 | seconds |
|
| 80 |
| `speaker_audio_text` | string | transcript of the speaker reference clip |
|
| 81 |
+
| `speaker_audio_source_enhanced` | binary (FLAC 24 kHz mono) | de-environment'd speaker reference: **MossFormer2-enhanced** for diarization sources; the **real clean-studio recording** for DDS |
|
| 82 |
| `text` | string | transcript of `speech` |
|
| 83 |
| `speech` | binary (FLAC 24 kHz mono) | target utterance, 3–15 s |
|
| 84 |
| `speech_duration` | float32 | seconds |
|
| 85 |
| `language` | string | `zh` / `en` / `auto` |
|
| 86 |
+
| `dataset` | string | `dds` / `m3sd` / `aishell4` / `msdwild` / `chime6` / `ami` / `alimeeting` |
|
| 87 |
+
| `conversation_id` | string | unique within the source dataset (for DDS: `{room}__{device}__{channel}`) |
|
| 88 |
+
| `speaker_id` | string | within-scene speaker label (for DDS: DAPS speaker, e.g. `f1`, `m8`) |
|
| 89 |
+
| `env_id` | string | acoustic-scene identifier (for DDS: `dds__{room}__{device}__{channel}`) |
|
| 90 |
| `text_source` | string | `original`, `asr`, or `mixed` |
|
| 91 |
+
| `asr_token_count` | int32 | Qwen3-ASR token count for `speech` (0 when `text_source=original`) |
|
| 92 |
| `asr_mean_logprob` | float32 | mean log-prob per token for `speech` |
|
| 93 |
|
| 94 |
## Source corpora
|
| 95 |
|
| 96 |
+
| dataset | hours (≈, this release) | language | transcripts |
|
| 97 |
| --- | ---: | --- | --- |
|
| 98 |
+
| **DDS — Device-Degraded Speech, DAPS portion** (Li & Yamagishi, 2021) | ~252 | en | ✅ DAPS scripts |
|
| 99 |
| **M3SD** (Wu et al., 2025) | 770 | zh / en mixed | ❌ → Qwen3-ASR |
|
| 100 |
| **AISHELL-4** (Fu et al., 2021) | 120 | zh | ✅ TextGrid |
|
| 101 |
| **MSDWILD** (Liu et al., 2022) | 80 | zh / en mixed | ❌ → Qwen3-ASR |
|
|
|
|
| 103 |
| **AMI** (SDM, diarizers-community) | ~100 | en | ❌ → Qwen3-ASR |
|
| 104 |
| **AliMeeting** (OpenSLR 119, far ch.0) | ~120 | zh | ✅ TextGrid |
|
| 105 |
|
| 106 |
+
**DDS** is single-speaker **read speech** (not a diarization corpus): 20 DAPS
|
| 107 |
+
speakers re-recorded across **9 rooms × 3 microphones × 6 positions** (162
|
| 108 |
+
acoustic conditions). For each condition the target `speech` and the
|
| 109 |
+
`environment_audio_source` (a *different* speaker, same room/mic/position) are the
|
| 110 |
+
**device-degraded** recordings, while `speaker_audio_source_enhanced` is the
|
| 111 |
+
matching **clean-studio** recording — a real enhancement ground truth rather than
|
| 112 |
+
a MossFormer2 estimate. All DDS text comes from the DAPS scripts
|
| 113 |
+
(`text_source = original`).
|
| 114 |
+
|
| 115 |
## Processing pipeline
|
| 116 |
|
| 117 |
Built with the streaming pipeline in
|
| 118 |
+
[`env-tts-data-pipeline`](https://github.com/ChristianYang/env-tts-data-pipeline).
|
| 119 |
+
|
| 120 |
+
Diarization sources — three parallel stages **download → process → upload**:
|
| 121 |
|
| 122 |
1. **download** — stream each source conversation (HF mirrors, OpenSLR tar
|
| 123 |
+
streams, etc.) into a bounded local cache; emit a JSON sentinel when ready.
|
|
|
|
| 124 |
2. **process** — resample to **24 kHz mono**, walk diarisation turns, emit
|
| 125 |
3–15 s `speech` slices with a same-speaker reference (≥2.5 s) and a
|
| 126 |
+
different-speaker environment slice (≥2.5 s). Missing/split transcripts are
|
| 127 |
+
re-labelled with **Qwen3-ASR-1.7B**. Snappy parquet shards (~800 rows / shard,
|
| 128 |
+
4 shards per HF commit group).
|
| 129 |
+
3. **upload** — `HfApi.upload_folder` per sealed group, resume-safe.
|
| 130 |
+
4. **enhance** (second pass) — **MossFormer2_SE_48K** on `speaker_audio_source`.
|
| 131 |
+
|
| 132 |
+
**DDS** uses a dedicated parallel-channel pass (`process-dds`): each (room,
|
| 133 |
+
device, position) condition is one acoustic scene; rows are assembled directly
|
| 134 |
+
from the parallel clean/degraded recordings, and `speaker_audio_source_enhanced`
|
| 135 |
+
is filled in-place with the real clean-studio clip (no second-pass MossFormer2).
|
|
|
|
|
|
|
|
|
|
| 136 |
|
| 137 |
## Licensing
|
| 138 |
|
| 139 |
+
Released under **CC-BY-NC-4.0** (non-commercial), inheriting the most restrictive
|
| 140 |
+
terms among sources. In particular:
|
| 141 |
|
| 142 |
+
- **DDS / DAPS** — CC-BY-NC-4.0 (non-commercial). *This is the binding term for
|
| 143 |
+
the whole release.*
|
| 144 |
- **M3SD** — academic / non-commercial research only.
|
| 145 |
- **MSDWILD** — X-LANCE research-only agreement.
|
| 146 |
- AISHELL-4 (Apache-2.0), CHiME-6 (CC-BY-SA-4.0), AMI, and AliMeeting carry
|
| 147 |
+
their respective open / research terms.
|
| 148 |
|
| 149 |
Redistributing extracted audio requires complying with each upstream licence.
|
| 150 |
|
|
|
|
| 153 |
Please cite the source papers when using this corpus:
|
| 154 |
|
| 155 |
```bibtex
|
| 156 |
+
@article{li2021dds,
|
| 157 |
+
title={DDS: A new device-degraded speech dataset for speech enhancement},
|
| 158 |
+
author={Li, Haoyu and Yamagishi, Junichi},
|
| 159 |
+
journal={arXiv preprint arXiv:2109.07931},
|
| 160 |
+
year={2021}
|
| 161 |
+
}
|
| 162 |
+
|
| 163 |
@article{wu2025m3sd,
|
| 164 |
title={M3SD: Multi-modal, Multi-scenario and Multi-language Speaker
|
| 165 |
Diarization Dataset},
|
|
|
|
| 193 |
```
|
| 194 |
|
| 195 |
ASR re-labelling uses [Qwen3-ASR-1.7B](https://huggingface.co/Qwen/Qwen3-ASR-1.7B).
|
| 196 |
+
Speaker enhancement uses MossFormer2 (ClearVoice). DDS is built on the
|
| 197 |
+
[DAPS](https://zenodo.org/records/4660670) dataset (Mysore, 2015).
|
| 198 |
|
| 199 |
## Loading
|
| 200 |
|
|
|
|
| 203 |
|
| 204 |
ds = load_dataset("ChristianYang/Env-TTS-Clean", split="train", streaming=True)
|
| 205 |
row = next(iter(ds))
|
| 206 |
+
print(row["text"], row["dataset"], row["speech_duration"])
|
|
|
|
| 207 |
# Audio columns decode automatically when accessed (24 kHz mono).
|
| 208 |
+
|
| 209 |
+
# Filter to a single source (e.g. the DDS read-speech rows):
|
| 210 |
+
dds = ds.filter(lambda r: r["dataset"] == "dds")
|
| 211 |
```
|
| 212 |
|
| 213 |
## Files on disk
|
| 214 |
|
| 215 |
```
|
| 216 |
data/
|
| 217 |
+
group_00000/ ... group_00122/ # group_00061–00122 are DDS
|
| 218 |
manifest.json
|
| 219 |
data_000000.parquet
|
|
|
|
|
|
|
|
|
|
| 220 |
...
|
| 221 |
```
|
| 222 |
|