There are several ways to improve the quality of live captions and translations in Timbra.
If you're using an HLS stream as your source, increasing the HLS segment size gives Timbra more context to produce higher quality captions.
We recommend using a segment size of 10 seconds. The minimum supported segment size is 6 seconds.
For Brightcove streams, see How to increase the HLS segment size in Brightcove.
Language profiles allow you to control how captions are generated. For example, you can configure:
These settings can also be configured on a per-language basis.
Custom dictionaries bias speech recognition towards specific words and phrases.
Custom dictionaries work with the primary and fallback transcription services. They are not available with the other services.
For live events, we recommend including:
This helps improve transcription accuracy, particularly for specialist vocabulary.