When Li Mu and Alex Smola's team developed the WildASR benchmark, they specifically selected MagicHub's natural full-duplex conversational speech data as a core data source. This not only represents recognition of our data quality by leading researchers, but also reinforces a fundamental insight: Real-world speech interaction must be learned from real human conversations, including hesitations, interruptions, and disfluencies.
Authentic human conversations shouldn't remain confined to research labs.
As part of our ongoing exploration of multilingual full-duplex speech data, MagicHub is once again making new training resources openly available: MAGICDATA-DUPLEX-SPANISH.
Why Spanish?
Spanish is one of the most widely spoken languages in the world and an essential language for multilingual Voice AI development.
Natural Spanish conversations feature rich interaction patterns, including rapid turn-taking, frequent backchannels, variations in intonation, spontaneous interruptions, and overlapping speech.
For example, a simple "sí," "claro," or "ajá" can signal agreement, encourage the other speaker to continue, or indicate an intention to take the next turn, depending on the conversational context.
These seemingly small details are precisely what enable full-duplex speech models to achieve natural, human-like interactions.
To move AI from simply understanding individual utterances to understanding entire conversations, we need more authentic data.
What Makes Our Dataset Different?
Real Human Conversations, Authentic Interactions.
All recordings come from real conversations between two speakers, preserving natural pauses, backchannels, interruptions, overlapping speech, and emotional variations.
Independent Dual-Track Recording, Preserving Original Interactions.
Each speaker is recorded on a separate audio track, allowing clearer analysis of individual speech and overlapping segments. This supports full-duplex models in learning real-time conversational behaviors.
Focused on High-Quality, Highly Interactive Content.
Through careful data selection and quality review, we retain conversation samples with meaningful natural interactions, minimizing the impact of low-quality audio on model training.
Designed for Real-World Voice AI Applications.
The dataset helps models learn immediate feedback, smooth turn transitions, and natural interruptions in Spanish conversations, providing a stronger data foundation for research into real-time speech interaction.
Who Is This Dataset For?
This dataset is suitable for research involving Spanish full-duplex speech models, multilingual Voice Agents, real-time voice interaction, turn-taking prediction, interruption detection, overlapping speech processing, and related fields.
Usage Notice: This dataset is released for non-commercial use only. It may be used for academic research, education, and non-commercial model training and evaluation. Commercial use is strictly prohibited. Please refer to the license terms on the dataset's official page for detailed usage requirements.
Why Magichub.com?
While there are many open-source data communities, dataset quality varies significantly. Many datasets appear impressive in scale, but once downloaded, they require extensive additional processing—inconsistent formats, non-standard annotations, misaligned audio and text, uncleaned noise, and missing metadata. Very few are truly ready for direct use in model training.
What sets Magichub.com apart is that all our datasets are designed to be ready to use out of the box. You don't need to spend additional time and resources on data cleaning, alignment, format conversion, or annotation correction. Every data sample undergoes rigorous quality inspection and standardization before release, allowing it to be used directly for training, evaluation, or product development.
For AI teams working under tight deadlines, this means saving weeks of data preprocessing and gaining faster iteration cycles and a more reliable starting point for experimentation.
MagicHub is becoming the foundation of real-world data for the Voice AI era.
