Hi guys,
I'm having a little bit of an issue with a small percentage of animations from a big chunk of batched anims. The vast majority are working great, but on a tiny amount, although the phoneme editor is showing what looks like appropriate phonemes for use with the words (for example the phrase 'Help me'), but the second word seems to get triggered far too late.
Does anyone know what might be causing this and what I can adjust to rectify it?
My setup contains the same poses as the slade character with some minor tuning of the mapping table.
Thanks in advance,
Matt.
Hi Matt,
Does the problem only happen when the file is analyzed in a batch process or does it also happen when you simply analyze the file individually? Is it possible for you to send us the audio file (and corresponding text you're using for analysis) to support at oc3ent.com?
- Jamie
Thanks for the reply Jamie. I'll do a test and see, I'll get back to you as soon as I can.
OK so I did some more analysis on a per file basis instead of batching but the results appear to be the same. I also tried this piece of audio with the slade character and I'm getting the same issue. I'll have a word with the powers that be here to see if we are able to give you some examples - as i am getting the same issue with Slade, can I just supply you with the wavs?
Matt,
Yes, all we need are a couple (one may be enough) of the problem .wavs as well as the text you are using for them during analysis. Are you seeing any warnings or errors printed in the console tab when this happens? Sometimes you will see warnings like 'audio level too low' or 'speech starts too soon' or something, followed by 'falling back to textless analysis'. You may be able to piece together what's going wrong by looking at the log in the console. My guess is you're getting one of these warnings and it's analyzing without the text, which is far less accurate.
- Jamie
Aha... the warning I'm getting is:
FxAnalysis: Fonix Warning: FnxCoreRunNode() returned warning 65538 (Inadequate leading silence. Please wait longer to speak.)
Is that likely to be the issue?
Yes, that's probably it -- it's probably defaulting to analysis with no text (you can confirm this if the phoneme / word bar in Studio has no words and only phonemes, though there should be a message to that effect in the logs somewhere). This warning means that the speech in the audio file starts immediately without a pause at the beginning. You can work around it by adding just a slight bit of silence to the beginning of these audio files. Alternatively, you can attempt to work around the issue by issuing the following command and then analyzing the problem file: set -n "a_detectspeech" -v "false" Note that this line is also in the autoexec.fxl file that ships with Studio but it is commented out by default; you could edit that autoexec.fxl file and uncomment the line and restart Studio and this setting will always be in effect.
- Jamie
That's odd in that case then, because the words are visible in the phoneme window...
I'll give it a shot with editing the fxl file and let you know how I get on.
Yeah, I think in that case at some point we changed the behavior so that the text can still go through. At any rate, we'd like to see the behavior here in-house to see if there's anything we can do to fix it; so if at all possible please do send along an example audio file with the text that we can take a look at.
- Jamie