Field notes on speech data & voice AI.
Practical writing on dataset quality, consent, provenance, alignment, multilingual collection, and the legal layer underneath modern speech models. Written for ML teams, data leads, and the legal counsel who has to sign off on what they buy.
Latest from the blog.
How to Build a Custom Voice AI Dataset From Scratch
A practical walkthrough of designing, recording, and validating a custom voice AI dataset for training or fine-tuning your own models.
Build vs Buy: Should You Create Your Own Speech Dataset or License One?
A clear-eyed comparison of building your own speech dataset versus licensing one. Cost, time, quality, and risk considered.
Collecting Voice Training Data: A Practical Guide for AI Teams
How to collect voice training data for AI: planning, recording, consent, transcription, and quality control. Start small, scale right.
Consent and Copyright in AI Training Data: What Teams Need in 2026
Consent and copyright in AI training data, explained for engineers and product leads. What the law expects, what to ask vendors, and how to ...
Conversational AI Training Data Explained — From Dialogue to Models
Conversational AI training data: what makes dialogue audio different, why it matters, and how to source it for your next voice or chat model...
Diarization for ASR Training: Why Speaker Labels Matter
Speaker diarization for ASR training: what it is, why it matters, and how to evaluate diarization quality before you buy a speech corpus.
How to Evaluate a Speech Dataset for AI Training Quality
How to evaluate a speech dataset before you buy: audio integrity, transcript accuracy, speaker diversity, consent, and a real training exper...
How to Fine-Tune Whisper on Your Own Audio Data
A practical guide to fine-tuning OpenAI Whisper on your own labeled audio. Data prep, training tips, evaluation, and common pitfalls.
GDPR and Voice Data: What AI Teams Need to Know
GDPR and voice data for AI training: what counts as personal data, what consent is required, and how to stay compliant when sourcing speech ...
What Makes a High-Quality Speech Dataset for AI Training
Quality is more than clean audio. Here are the dimensions that actually matter when you are building or buying a speech dataset for AI train...
How Much Audio Do You Need to Train a Speech Model?
How much audio do you need to train a speech model? Real numbers for ASR, TTS, and conversational AI — plus why hours alone do not tell the ...
How to License Speech Data for AI Training in 2026
Learn how to license speech data for AI training in 2026 — sourcing, consent, formats, pricing, and what to ask before you buy. Get a quote.
How Podcasters Can Monetize Their Audio with AI Licensing
How podcasters can monetize their back catalog by licensing audio for AI training. What the deal looks like, how much it pays, and how to st...
Multilingual Speech Datasets for AI: Coverage, Cost, and Pitfalls
Multilingual speech datasets for AI training: how to source, how to evaluate, and what really matters when shipping in more than one languag...
Phonetic Balance in TTS Training Data: Why It Matters and How to Achieve It
A phonetically balanced TTS dataset trains a more natural voice with less data. Here is how to design and validate balance in your corpus.
Why Podcast Audio Is Ideal for AI Training Datasets
Podcast audio combines studio quality with natural conversation. Here is why it has become a favorite source for speech model training.
Why Provenance Matters in AI Training Data — and How to Prove It
Provenance in AI training data is now a procurement requirement. Learn what provenance means, what to document, and how to prove it to custo...
Speaker Diversity in AI Training Data: Why It Matters and How to Get It
Speaker diversity is one of the strongest predictors of how well a speech model generalizes. Here is how to think about it and how to actual...
How Much Does Speech Training Data Cost in 2026?
Speech training data cost in 2026: per-hour rates, what affects pricing, and how to budget. ASR, TTS, and conversational data — all explaine...
Studio vs In-the-Wild Audio: Which Is Better for Training Speech Models?
Studio audio is cleaner, but in-the-wild audio teaches models to handle the real world. Here is how to balance both in your training set.
Synthetic vs Real Speech Data for AI Training: Which Wins?
Synthetic vs real speech data for AI training: when to use each, where synthetic falls short, and the hybrid approach most teams settle on.
Transcript Alignment for ASR Training: Why It Matters and How to Do It Right
Alignment turns a transcript into useful labels. Here is why it matters for ASR training and how to produce alignments that hold up.
TTS Dataset Requirements: What Makes Voice Synthesis Data Train Well
TTS dataset requirements explained: speakers, recording quality, phonetic balance, scripts, and metadata. Build or buy a corpus that synthes...
Legal Considerations for Voice Cloning Datasets
Voice cloning sits in some of the most active legal territory in AI. Here is what dataset builders need to know about consent, rights, and r...
What Is ASR Training Data? A Plain-English Guide for AI Teams
ASR training data explained: what it is, how it is labeled, what to look for in a dataset, and why provenance matters in 2026. Get a sample.