Better lip syncing

[thefourwings]

Thanks in advance for your help...

 

I'm integrating voice and lip sync files and it's working out fine so far, except that the lip sync itself doesn't look very good. I know F3's lip syncing isn't all that great to begin with, but mine seems to have turned out even worse than the vanilla syncs, and I'm wondering if I've done something wrong or maybe that's just as good as it gets from the construction set. Here's a quick peak at some of the lines so far: http://www.thefourwings.com/IRCmod/IRC%20m...%20Computer.m4v

 

 

Here's what I'm doing to make/integrate the files:

1.) move the "temp" dialogue wav into the TES construction set

2.) type out the line of dialogue and hit save (generates named wav and lip)

3.) rename new wav "temp" and move to F3 folder

4.) find the right line of dialogue and hit save (generates named wav)

5.) find the new wav, copy the file name and use it to rename the lip file, then move that file to the F3 folder

 

 

As you can see from the movie, it does work - just doesn't look all that good. Can anyone see something I'm doing wrong, or how I could make it look more accurate - or even if someone understands how the construction set goes about generating the lip (is it based on the text i type in and/or does it analyze the sound waves of the wav file?)

 

Thanks again for any help. I've gone to the trouble of writing & getting voice actors for almost 400 lines of dialogue and it seems a shame for it to look so poorly synced.

[BadPenney]

I can't see your video online and I'm too paranoid to download it, but:

 

Sometimes I have to delete the LIP file and generate a new one to improve the lipsynch.

 

I use a combination of the GECK, TES Construction Set, Sony SoundForge and MS Word to make audio dialogue. The more distinct and clear that your WAVs sound then the more likely that you will get good lip files. I use SoundForge to check the waveforms of the dialogue that I am using. and enhance them occasionally.

 

In some cases, the sounds are too indistinct to get good LIP files, so I record myself repeating the line with exaggerated emphasis on the problem areas. Vowell sounds especially seem to get muddled. After recording the new WAV, I cut and paste the exaggeratedly clear and loud portions into a copy of the original line of dialogue with SoundForge, being careful to maintain the timebase of the original recording. Then I use that to create a new LIP file and combine that with the original WAV. I am often able to get a better lip synch that way.

 

I think that the Construction Set mainly uses the audio file to generate the LIP, but sometimes it seemed that if I didn't write the dialogue out that the LIP would be bad, so I got into the habit of using an exactly written line, just in case. I need one for the final version anyway.

 

In case that you are interested in the sequence that I used for generating audio dialogue for Return to Shady Sands it was like this:

 

Open folders for the source sound files, the TES sounds (I always use Imperial male to keep it simple), and a folder for the target NPC in Fallout.

 

Open Word, SoundForge, GECK and Construction Set. Load the WAV into SoundForge, listen to it and type out the phrase in Word. Use spell check if necessary.

 

Create an Imperial Male NPC in the CS. (One NPC can do all voices regardless of gender or sound differences. It makes finding the files simpler.) Start a quest in the CS and create new dialogue. Cut and paste the line from Word into the CS. Create the condition for the Imperial Male NPC that you made. Paste the WAV into the Imperial Male folder. Click on the line of dialogue to generate the file name. Copy that name onto the WAV file in the Imperial Male folder. Close the line of dialogue in the CS and open it again. If the WAV shows up in the list with a Y click on it to listen and then click "Generate Lip". Your Imperial Male folder should now have a WAV and a LIP with the same file name.

 

After creating the LIP file, I like to use SoundForge to create an OGG file from the WAV. It is the same type that Fallout 3 uses, is near CD quality, and uses less memory.

 

Copy both the OGG or WAV file and the LIP into the target FO3 NPC dialogue folder. Open the target FO3 quest and start new dialogue for the target NPC. Paste the same line of dialogue from Word into the new dialogue box. Click on the dialogue line to generate a file name. Highlight the filename and type CTRL + C to copy it. Highlight both the OGG and the LIP in the target FO3 NPC folder, select rename and type CTRL + V to paste the assigned GECK filename onto both of the files. Close the dialogue box in the GECK and open it again to check the sound and LIP are recognized. SAVE.

 

That was just my way of doing the process, but it worked for me. I also like to clear out the CS Imperial Male folder occasionally so that it doesn't become hard to distinguish which file that I am working on, but keep backup folders for each NPC that I am making dialogue for and save all associated files in them at least until I am confident that the backups will no longer be needed.

 

The final step is of course checking the line of dialogue in the game to see if the lip synch is good. If not, try regenerating the lip. Sometimes that is all that you need to do. If it still doesn't look good. making the WAV louder or filtering it before generating the LIP can sometimes help. You don't have to use the file that actually generated the LIP in the game. As long as the time base is the same then you can pair the LIP with the sound that you like best.

[thefourwings]

Thanks for the tips. Not going to rerecord an exaggerated phonetics version for all 400 lines, but will probably do that for my next mod, which is also dialogue-heavy. At least I have a better idea now of how to structure my script so that it's more lip sync conducive. Most of the lines look pretty decent (considering even the vanilla lip syncs are all that great) so I think I'll just put up with it and then when I'm beta testing, set aside the lines that are especially bad to rework them. Thanks again,