# IndecalAI blog: speech data and AI training data

Source: https://indecalai.com/blog/

Practical posts for teams building speech and AI models: how much data you need, where to get it, and what to check before you buy. From the people who collect it.

## Posts

- [How many hours of speech data do you need to fine-tune Whisper?](https://indecalai.com/blog/how-many-hours-of-speech-data-to-fine-tune-whisper/) (2026-10-27, Speech data): How much audio you need to fine-tune Whisper or another ASR model: what the Whisper paper shows, rough starting points by goal, and how to size a speech dataset without guessing. Markdown copy: https://indecalai.com/blog/how-many-hours-of-speech-data-to-fine-tune-whisper/index.html.md
- [Off-the-shelf or custom speech data: how to choose](https://indecalai.com/blog/ots-vs-custom-speech-data/) (2026-10-27, Datasets): When to license an off-the-shelf speech dataset and when to commission a custom collection: a side-by-side comparison, five questions that decide it, and what to check in any licence. Markdown copy: https://indecalai.com/blog/ots-vs-custom-speech-data/index.html.md

## Guides

- [Speech dataset specification: what to put in your collection brief](https://indecalai.com/guides/speech-data-collection-specs.html): Scripted or spontaneous speech, who should be speaking, recording conditions and metadata, plus a brief template you can copy.
- [Transcription for AI training data, versus for people to read](https://indecalai.com/guides/transcription-for-ai-vs-people.html): Verbatim vs clean text, timestamps, speaker labels, and how to decide which one your project needs.
- [How to choose a data annotation vendor: the questions worth asking](https://indecalai.com/guides/choosing-a-data-annotation-vendor.html): The questions worth asking a vendor, what good answers sound like, and how we answer them ourselves.
- [What makes data annotation quality good, or bad](https://indecalai.com/guides/data-annotation-quality.html): Guideline clarity, inter-annotator agreement, how edge cases get resolved, and the review layers that catch mistakes.
- [Synthetic data vs real data: when to use each](https://indecalai.com/guides/synthetic-vs-real-data.html): What each kind of data is good at, where synthetic data goes wrong, and how to test a model that uses both.
- [How to plan an egocentric video data collection](https://indecalai.com/guides/egocentric-video-data-collection.html): The decisions behind a first-person video dataset, from device and mount to privacy, labels and checks, with a brief template to copy.

RSS feed: https://indecalai.com/blog/feed.xml
Talk to our team: https://indecalai.com/contact.html