database menu talk clips

[royal]

I've been thinking about the possibility of making a folder tree for talk clips that the database menu uses.  Let's say .rockbox/dbtalk where rockbox would use the .talk file corresponding to the database menu option if it found one (.talk file, that is).  Although I currently have no idea how hard this would be, it sounds like it wouldn't be too complicated.
I understand it would make it cumbersome to keep up with the .talk clips should one change the tagnavi.custom file considerably, but it seems worth it.
I'd like to know some developer's opinion on this subject, and if no desire can be found to establish this feature, maybe a few pointers to a newbie...

[pondlife]

It might be simplest to have .talk files named after the text that they voice, all in a single directory. e.g.

Album.talk
Album Artist.talk
...
2009.talk
...
Elvis Costello.talk
...
etc. etc.

The run-time engine would then just need to do a simple phrase lookup.

The major work would be automatic generation of the clips, but I suppose that could be done as part of an external database build utility (something which has been talked about a lot, not sure if it exists).

(OR the TTS project might get off the ground...)

pondlife

[Llorean]

Not only the simplest, but (to an extent) the only reasonable way considering the range of tags you can have the database filtered by.

There'd also need to be some way to have it be able to use multiple strings. For example, if you had it displaying both the year tag and the album tag, it would need to be able to recognize that it's looking for two separate strings unless your voice building program took a tagnavi as input and read all strings that the current database would generate. That would result in a lot of redundant strings, but would at least improve accuracy if there were any words/characters in the strings that didn't come directly from the tags.