Transcribe and inde...
Clear all

Transcribe and index dialog

3 Posts
3 Users
Topic starter

Transcribe the text/speech found in the media and make it searchable. This would revolutionize the process of finding alts and especially, cheats, in production audio -- or any other dialog/speech editing context.


For example, imagine having instant access to all occasions where a character said a certain word -- or syllable even -- outside of the given scene; ready to preview and spot... Would save many hours if not days on some projects.


Any chance something like has been considered or might be in the works?

2 Answers

Nothing along these lines planned, or at least nothing like this has ever been requested or discussed which doesn't mean it's not a good idea, just not something people are really looking for in a sample manager I guess.

There are software solutions out there that can generate test from speech but there would be a few issues to incorporate something like this into BaseHead

First would be the service needed to do this work and the cost vs the demand for the feature. Next would be the need to store a large amount of data in the database. 

It's a valid request, just not something I think is in consideration at least for the next year or two. 


Absentia DX just released a beta transcribe function.  If anything, it would be interesting to add support for importing their transcribed spreadsheet along with the production dialog to make it searchable. 

makarushka Topic starter 02/01/2023 10:46 am

Yes! I have been involved in some testing of the Absentia DX from a while back. I just played with the current iteration of the transcribe tool a few days ago and had sent my feedback to the developers. It is already very usable: the transcribe function can embed metadata that creates markers/regions that do show up in Basehead -- see the screenshot. So you can already scan the file(s) for specific text/speech much quicker. I only wish the channels had their names displayed on the waveform horizontally -- at the left or rightmost end of the waveform; then you'd know immediately which character the text belonged to.


Absentia tool itself is for now very slow to transcribe and embed -- that is perhaps its biggest issue. There are also other, smaller, issues, which no doubt will be much easier to fix. But as it stands already, this is a huge help.