Datasets:
The dataset viewer is not available for this subset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
~~~~~~~~~~~~~~~~~~~~~~~~~^
StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 84, in _split_generators
raise ValueError(
...<2 lines>...
)
ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
~~~~~~~~~~~~~~~~~~~~~~~^
path=dataset,
^^^^^^^^^^^^^
config_name=config,
^^^^^^^^^^^^^^^^^^^
token=hf_token,
^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
path,
...<6 lines>...
**config_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
MotionSpeech: A Co-speech Motion and Speech Dataset
MotionSpeech is an aligned multimodal dataset of 34,950 paired clips (~12.9 s each) of synchronized speech audio and full-body SMPL-X motion, with multi-layer textual annotations. It is assembled from two public sources and re-segmented, re-annotated and quality-filtered for training text-driven joint motion+speech generative models.
| Source | Clips | Notes |
|---|---|---|
| BEAT2 (English) | 17,907 | scripted monologue gestures, 30 speakers |
| Seamless Interaction (improvised segments) | 17,043 | improvised English dyadic interaction |
Download and extraction
This release is distributed as lossless TAR archives (plain .tar or gzip-compressed .tar.gz) under data/, grouped by the
official train, validation, and test splits (up to 1,000 samples per shard).
Each sample retains its original WAV, SMPL-X NPZ, and caption TXT bytes and filenames.
Extracting all shards reconstructs the directory structure shown below.
pip install -U huggingface_hub
hf download JunDong-dev/motionspeech --repo-type dataset --local-dir ./motionspeech_data_opensource
cd motionspeech_data_opensource
sha256sum -c SHA256SUMS
for archive in data/*.tar*; do
tar -xf "$archive"
done
split/ contains the original split lists. metadata/files.jsonl records the
relative path, split, byte size, and SHA-256 of every sample file;
metadata/release.json lists all shards and sample counts.
Directory structure
motionspeech_data_opensource/
βββ README.md this file
βββ split/ official splits (sample IDs, one per line)
β βββ train.txt 27,957
β βββ val.txt 3,501
β βββ test.txt 3,492
βββ audio_wav/ <sample_id>.wav mono 16-bit PCM
βββ motion_smplx/ <sample_id>.npz SMPL-X + FLAME pose sequence
βββ caption/ <sample_id>.txt multi-layer annotation (see below)
All 34,950 annotated clips have matching audio, motion, and caption files with the same sample ID.
Sample ID convention
Two naming patterns, inherited from the sources:
- BEAT2:
{speaker_id}_{name}_{take}_{session}_{seg}_{start_frame}_{end_frame}e.g.10_kieks_0_10_10_1256_1642 - Seamless:
V{video}_S{session}_I{interaction}_P{participant}_{start_frame}_{end_frame}e.g.V03_S1399_I00000495_P4044_7194_7580
All files of one clip share the same ID (only the extension differs).
Audio (audio_wav/)
- WAV, mono, 16-bit PCM.
- BEAT2 clips are 16 kHz; Seamless clips are 48 kHz (native rates of the sources are preserved β resample on load if a uniform rate is needed).
- Each clip is ~12.9 s, aligned 1:1 in time with its motion file.
Motion (motion_smplx/)
Each .npz contains one clip at 30 fps (387 frames β 12.9 s):
| Key | Shape | Description |
|---|---|---|
poses |
(T, 165) | SMPL-X joint rotations in axis-angle (55 joints Γ 3) |
trans |
(T, 3) | global root translation (meters) |
expressions |
(T, 100) | FLAME facial-expression coefficients |
betas |
(300,) | SMPL-X body-shape coefficients |
model |
str | model identifier (smplx+flame) |
gender |
str | male / female |
mocap_frame_rate |
int | capture frame rate (30) |
Load with numpy.load(path); drive SMPL-X+FLAME with the poses/trans/betas
to obtain meshes. (For 6D-rotation training features used in our baselines,
poses can be converted 3-axis-angle β 6D per joint.)
Captions (caption/)
Each .txt holds four tagged layers concatenated in one line:
<L1>{one-sentence summary of the gesture}</L1>
<L2>{step-by-step detailed gesture description}</L2>
<S>{final speech transcript}</E>
<AUDCAP>{speaker voice description: gender, timbre, tone, pace}</AUDCAP>
The line breaks above are for readability. Every caption uses the exact order
<L1>...</L1><L2>...</L2><S>...</E><AUDCAP>...</AUDCAP>, with all four
sections nonempty. The speech section closes with </E>.
Example (truncated):
<L1>The hands stay near the abdomen while the body shifts laterally, then both
palms open slightly outward.</L1>
<L2>First, both elbows remain bent and the hands are held apart in front of the
abdomen with gently curved fingers. Next, the hands move closer together ...
Finally, the hands remain low and open ...</L2>
<S>I would prefer to choose a major that is easy to find a good job in the
future, like finance or marketing ...</E>
<AUDCAP>Female voice, clear and steady, speaking in an informative tone.</AUDCAP>
L1/L2describe co-speech body gestures at two granularities.Sis the final speech transcript of the paired audio, using the accepted Whisper (whisper-large-v3-turbo) transcription or the final human review decision for both BEAT2 and Seamless clips. Human decisions may accept the new transcription, retain the original, or provide a manual correction. Capitalization and punctuation are preserved; automatic transcriptions should not be treated as human-verified ground truth.AUDCAPis a compact speaker-voice caption (gender / timbre / tone / pace) usable as a conditioning prompt for voice characteristics.
Intended use
Text-driven / phoneme-driven joint generation of co-speech gesture and speech (TTS-style audio + gesture avatars), gesture-speech synchronization research, and captioning of multimodal speech+motion streams.
License & attribution
This dataset is a re-segmentation and re-annotation of two public corpora. Users must comply with the original licenses of BEAT2 and Seamless Interaction, and cite both sources in any derived work.
Citation
If you use MotionSpeech in your research, please cite the following paper in addition to the original source datasets:
DuoFlow: Towards Synchronized Speech and Motion Generation with Fine-Grained Control
@misc{dong2026duoflow,
title = {{DuoFlow}: Towards Synchronized Speech and Motion Generation with Fine-Grained Control},
author = {Jun Dong and Ronghui Li and Zhongyuan Hu and Yunqi Zhao and Lijie Guo and Ruizhang Xu and Di Kang and Linchao Bao and Xiu Li},
year = {2026},
eprint = {2611.21530},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2611.21530}
}
- Downloads last month
- 53