How to improve write to file speed (in v3.x)?

[caricatur]

I remember Textaloud 2.x had an option to change write speed. I've been using it on my netbook and just got a new PC. On my netbook, cpu load is btw 50-70% while writing txt to mp3. On my other PC it is only 11-14% and it only writes about 1.5 times faster despite the fact that it is 10x faster than my netbook.

I spend alot of time converting my books to audio to listen to while I travel or even around the house. Is there a way to have text aloud output the file faster? I looked all over the place and I couldn't find a write to file speed setting, or anything that seems to indicate an increase in processing power. I would like the faster machine to write the file at a higher speed. Is this possible with 3.x?

What are my options? I don't think there is a way to write more than one file at a time...

Thanks

[Jim Bretti]

The write to file speed option you're referring to only works with older sapi4 voices, it is ignored in Sapi5. In TA3, we decided to set this at fairly high value and removed the option since it was causing some confusion. If you're using sapi4 voices and you need to configure this let me know.

The only thing I can think of is to look at the Advanced Settings dialog (under the Tools menu). The third option on the Settings panel is called "Raise Processing Priority when writing audio files", and I'm pretty sure it defaults to disabled (unchecked). If the setting is disabled, try enabling it and see if it makes any difference. There are no other Sapi5 options I'm aware that we can use to control this.

Let me now what your results are with the processing priority setting, I'm curious to know if that has any effect.

[caricatur]

I tested it with that option checked and unchecked. I also set the priority to high, but neither of these seemed to have a noticeable effect.

Thanks for the reply. I'm used to the program taking up a considerable amount of processing power--under my current setup it isn't noticeable. It would be nice if it could be configured to output files faster, but the ability to write directly to mp3 is a huge improvement over v2.x.

Being able to set the voice speed to more than 30 would also be nice. I usually play the files back at around 1.5x. or convert it for playback on my media player. Would it be possible to change the limit in future versions, or is this due to how TTS voices work?

[Jim Bretti]

The voice speed setting depends on the voice. Sapi5 requires the voice handle a range from -10 to 10. We expand that range in TextAloud to -30 to 30, only because we noticed that some voices handle it. Some voices max out at 10 though. Let me know which voice(s) you're using and I'll do some checking.

[caricatur]

That makes sense. When I did tests earlier, there was no difference between 10 and 30x. I thought it due to some strange bug or fuzzy formula. 7x is 95% the speed of 30x, but there is a dramatic difference between 4x and 7x. I believe both voices I generally use max at 10x. Is there a way to look up which voices handle 30x - how would I find out that information before purchase?

Currently, playback at 2x ~600 wpm) is near undecipherable. I've tried adding very small pauses in between each word to see if I can hear each word distinctly, since at that speed they all start to meld together. Is there a minimum I can apply? I believe I used a value like 0.0025 delay after each word, but it just ended up sounding stilted and unnatural. What is the minimum that I can use (theoretically) that TextAloud will handle differently?

Currently I listen to most files at 1.5 (~450 wpm) At 540 wpm, I need to concentrate a little more to absorb the information. That's still below moderate levels of speed reading. I stick with TA since it doesn't slow down and I can usually listen from 300wpm comfortably and 450wpm) if I'm not doing anything too distracting. I can understand at 520 wpm, but my recollection starts to suffer at that rate.

I generally read around 300-450 wpm when I haven't practiced n a while (my state currently) I feel the trade-off is usually worth it, but knowing that with minimum practice I can maintain 750 wpm easily (that is not too fast if you stop sub-vocalizing- technically, it still happens at 350-450. I believe most people can't read above 450 if they sub-vocalize.

TA has been my #1 productivity tool. If it weren't for it, I wouldn't "read" anywhere near as much. I have no interest in abandoning it, but its weakness is it's (relatively) low speed and the fact that it only reads one line at a time. I'm primarily an auditory learner, but I haven't trained my ears to read at the speeds my eyes can. If I read multiple lines as I've been able to do when I actually practiced, TA starts to feel really slow by comparison. A slight preference for sound, and shaky reading skill keep my glued to TA for now.

It would be nice to find a voice that can perform clearly at faster speeds. I don't know if I should work with TA timings or shop for a voice that works decently at those speeds. Would you know if an accented or unaccented voice works better? Real people use unaccented since it decreases the amount of movement required, but that is not relevant for machine speech.

I would really like to know if there are some speed reading voices out there that remain clear at 600+ wpm, or are designed for that. Are there any voices like Steve Woodward (637wpm) or Fran Capo (603)? My preferred voice ignores the punctuation filter. so I've been using my usually neglected voice which will follow the TA settings.

If you have advice on how I can research more technical details of voices, I would appreciate it. Most websites don't advertise the details, and I've only bought voices at sites like yours that offer voices and bundles. TA is by far the best TTS app in my opinion.

Thanks for the hard work. Take care

[Jim Bretti]

AT&T and Acapela voices handle speeds > 10. If you're interested, send me an email (jim@nextup.com) letting me know which voices you're interested along with a small text sample. I'll generate some sample mp3's so you can get an idea what they sound like.

Also, from what I've seen, Ivona voices seem to do pretty well also with the high speeds. We don't currently sell these voices but you can download eval copies of their voices at http://www.ivona.com/