Will someone post examples on processing restoration?

[rocky12]

Hello,

Will someone please post examples on processing restoration? Will someone also explain the lines of the script? Thank you.

[redwudz]

Script for what? Which program are you asking about?
What format? What type of restoration?

If it's AVISynth, start here: http://avisynth.org/mediawiki/Main_Page

You would have to provide a lot more information before anyone could give you any advice.

Describe what you want to accomplish.

[lordsmurf]

Original post:
I don't use or need AVISynth. I use hardware, VirtualDub, TMPGEnc Plus, Premiere and some other specialties (anti-shake video, audio, etc) Scripting would piss me off, make an already lengthy process take a hell of a lot longer than is required.

Update: I've started to add more Avisynth as needed, but only when absolutely required.

[redwudz]

I use VD most of the time also. I really like the 'gradation curves' filter to adjust contrast and brightness levels. It's saved a couple of flat looking videos. I used the 'levels' filter more in the past, but 'gradation curves' does what I want. And I use ColorMill a lot. It has most all needed adjustments in one filter.

I've used De-Shaker a few times with good results. And I still use Audacity and ffmpegGUI for audio most of the time. Just use to them.

On a really noisy video, I've had good luck with the Neat Video filter, though it's a bit expensive. I could probably use a combination of other filters to accomplish the same thing, but I find it quick and easy. I don't mind a slightly 'plastic' looking video if the original was mostly noise anyway. But I keep the original in case I want to spend a bit more time with it.

I use AVIsynth mostly for feeding RM(VB) and WMV to VirtualDub. Just haven't taken the time to understand it for other uses.

[lordsmurf]

I have to use Colormill for the first time here this weekend.

I'll try gradation curves again, since you like it. I'm still using levels, stuck in a rut, I guess. But hey, seems to work when I need it. Generally I fix those problems in a proc amp anyway. Only need these filters are already-digital source somebody else screwed up!

[jagabo]

I use the FixEverythingThat'sWrongWithThisVideo() filter. Works perfectly every time.

[themaster1]

Restoring low-quality Youtube/similar videos (avs scripts)

[poisondeathray]

Some nice work by VideoFred over at Doom9

The script is quite complex, but you can learn lots from it. I don't understand all of it (yet). If you are polite and respectful, I'm sure you can ask specific questions in that thread

You can find it in this post
http://forum.doom9.org/showthread.php?t=144271

http://www.vimeo.com/2823934

http://www.vimeo.com/2823934

# film restoring script by videoFred.
# denoising, resizing, stabilising, sharpening, auto-levels and auto-white balance.


film="F:\002_dodcaps_tebewaren\privé 64\privé_64T_0014.avi" # source clip, you must specify the full path here

short="L"  # L=long clip  S=short clip try it!

result="resultS3" # specify the wanted output here 

trim_begin=6  trim_end=6   play_speed=18.75   #trim frames and play speed (PAL: 16.6666 or 18.75)

saturation=1.0   #saturation

X=4 gamma=0.7  # X is a special parameter for reducing the autolevels effect

black_level=0  white_level=255 output_black=0  output_white=255 # manual levels, when returning result4

AGC_max_sat=2  AGC_max_gain=1.0  #parameters of HDRAGC filter, improves colors and shadows
 
blue=-4  red=2  #manual color adjustment, when returning result2. Values can be positive or negative

denoising_strenght=800  #denoising level of first denoiser: MVDegrainMulti() 
denoising_frames= 4  #number of frames for averaging (forwards and backwards) 3 is a good start value
block_size= 16  #block size of MVDegrainMulti()
block_over= 4  #block overlapping of MVDegrainMulti()

temp_radius=20  temp_luma=6  temp_chroma=6 #second denoiser: TemporalSoften

grain_luma=10  grain_chroma=10
# this will add some digital grain to the final result, set it to zero if you do not want it.

LSF_sharp_ness=250 LSF_radi_us=3  LSF_sub=1.5 #first sharpening parameters (LimitedSharpenFaster) sub=subsampling

USM_sharp_ness=40    USM_radi_us=2  USM_thres_hold=0  #second sharpening parameters (UnsharpMask)

USM_sharp_ness2=20    USM_radi_us2=1  USM_thres_hold2=0  #third sharpening parameters (UnsharpMask)

maxstab=60  #maximum values for the stabiliser (in pixels) 20 is a good start value

est_left=20   est_top=20  est_right=20  est_bottom=20  est_cont=0.8  #crop values for special Estimate clip
CLeft=30  CTop=30  CRight=30  CBottom=30  #crop values after Depan and before final resizing (40,30,40,30)

W=720  H=576 #final size from the returned clip

bord_left=0  bord_top=0  bord_right=0  bord_bot=0
#you can add black borders after resizing, final size is then size + borders!!


# End variables, begin script
#====================================================================================================


SetMemoryMax(1024)  #set this to 1/3 of the available memory


Loadplugin("Depan.dll")
LoadPlugin("DepanEstimate.dll")
Loadplugin("removegrain.dll")
LoadPlugin("AGC.dll")
LoadPlugin("MVTools.dll")
Loadplugin("mt_masktools.dll")
LoadPlugin("MaskTools.dll")
Loadplugin("warpsharp.dll")
LoadPlugIn("LimitedSupport_09Jan06B.dll")
LoadPlugin("MT.dll")
LoadPlugin("autolevels.dll")
LoadPlugin("AddGrainC.dll")
Import("LimitedSharpenFaster.avs")


SetMTMode(5)


source1= Avisource(film).assumefps(play_speed).trim(trim_begin,0).converttoYV12()
end= source1.framecount()
end2= end-trim_end
frames=end+trim_begin

skip= end2/5
skip0=skip+3
skipend= 3

skipend1= skip0+3

skip2= skipend1+skip
skipend2= skip2+3

skip3= skipend2+skip
skipend3= skip3+3

skip4= skipend3+skip
skipend4= skip4+3

skip5= end2-3

L= trim(source1,0,end2)
LS= trim(source1,0,end2).scriptclip("""subtitle("frame  "+string(trim_begin+current_frame)\
+"  from  "+string(frames)+"   trim_begin-"+string(trim_begin)+"   trim_end-"\
+string(trim_end),x=100,y=60,size=32)""")

sourceT1= trim(source1,0,skipend).scriptclip("""subtitle("frame  "+string(trim_begin+current_frame)\
+"  from  "+string(frames)+"   trim_begin-"+string(trim_begin)+"   trim_end-"+string(trim_end)\
+"  steps: "+string(skip),x=100,y=60,size=32)""")

sourceT2= trim(source1,skip0,skipend1).scriptclip("""subtitle("frame  "+string(trim_begin+skip0+current_frame)\
+"  from  "+string(frames)+"   trim_begin-"+string(trim_begin)+"   trim_end-"+string(trim_end)+"  steps: "\
+string(skip),x=100,y=60,size=32)""")

sourceT3= trim(source1,skip2,skipend2).scriptclip("""subtitle("frame  "+string(trim_begin+skip2+current_frame)\
+"  from  "+string(frames)+"   trim_begin-"+string(trim_begin)+"   trim_end-"+string(trim_end)+"  steps:\
 "+string(skip),x=100,y=60,size=32)""")

sourceT4= trim(source1,skip3,skipend3).scriptclip("""subtitle("frame  "+string(trim_begin+skip3+current_frame)\
+"  from  "+string(frames)+"   trim_begin-"+string(trim_begin)+"   trim_end-"+string(trim_end)+\
"  steps: "+string(skip),x=100,y=60,size=32)""")

sourceT5= trim(source1,skip4,skipend4).scriptclip("""subtitle("frame  "+string(trim_begin+skip4+current_frame)\
+"  from  "+string(frames)+"   trim_begin-"+string(trim_begin)+"   trim_end-"+string(trim_end)+\
"  steps: "+string(skip),x=100,y=60,size=32)""")

sourceT6= trim(source1,skip5,end2).scriptclip("""subtitle("frame  "+string(trim_begin+skip5+current_frame)\
+"  from  "+string(frames)+"   trim_begin-"+string(trim_begin)+"   trim_end-"+string(trim_end)+\
"  steps: "+string(skip),x=100,y=60,size=32)""")

SS= sourceT1+sourceT2+sourceT3+sourceT4+sourceT5+sourceT6


sourceT10= trim(source1,0,skipend)
sourceT20= trim(source1,skip0,skipend1)
sourceT30= trim(source1,skip2,skipend2)
sourceT40= trim(source1,skip3,skipend3)
sourceT50= trim(source1,skip4,skipend4)
sourceT60= trim(source1,skip5,end2)

S= sourceT10+sourceT20+sourceT30+sourceT40+sourceT50+sourceT60


stab_reference= eval(short).crop(est_left,est_top,-est_right,-est_bottom)\
.tweak(cont=est_cont).binarize(threshold=80).greyscale().invert()

mdata=DePanEstimate(stab_reference,trust=1.0,dxmax=maxstab,dymax=maxstab)
stab=DePanStabilize(eval(short),data=mdata,cutoff=0.5,dxmax=maxstab,dymax=maxstab,method=1,mirror=15)
stab2= stab.crop(CLeft,CTop,-CRight,-CBottom).tweak(sat=saturation)
stab3=DePanStabilize(eval(short),data=mdata,cutoff=0.5,dxmax=maxstab,dymax=maxstab,method=1,info=true)

WS= width(stab)
HS= height(stab)
stab4= stab3.addborders(10,10,10,10,$B1B1B1).Lanczos4Resize(WS,HS)


vectors= stab2.MVAnalyseMulti(refframes=denoising_frames, pel=2, blksize=block_size, overlap=block_over, idx=1)
denoised= stab2.MVDegrainMulti(vectors, thSAD=denoising_strenght, SadMode=1, idx=1).tweak(sat=saturation)




leveled= denoised.HDRAGC(coef_gain=2.0,max_gain=AGC_max_gain,min_gain=0.5,max_sat=AGC_max_sat,shadows=true)

sharp1=limitedSharpenFaster(leveled,smode=1,strength=LSF_sharp_ness,overshoot=50,\
radius=LSF_radi_us, ss_X=LSF_sub, SS_Y=LSF_sub, dest_x=W,dest_y=H)

sharp2= unsharpmask(sharp1,USM_sharp_ness,USM_radi_us,USM_thres_hold)
sharpX= unsharpmask(sharp2,USM_sharp_ness2,USM_radi_us2,USM_thres_hold2)

sharp3= sharpX.TemporalSoften(temp_radius,temp_luma,temp_chroma,15,2).addgrainC(grain_luma,grain_chroma,0.2,0.2,5)




#backward_vectors = sharp3.MVAnalyse(isb = true,truemotion=true,idx=2)
#forward_vectors = sharp3.MVAnalyse(isb = false,truemotion=true,idx=2)
#frameclip=MVFlowFps(sharp3,backward_vectors, forward_vectors, num=25, den=1, ml=100, idx=2)



result1= sharp3.addborders(X,0,0,0,$FFFFFF).levels(0,gamma,255,0,255).autolevels().coloryuv(autowhite=true)\
.crop(X,0,-0,-0).addborders(bord_left, bord_top, bord_right, bord_bot)

result2= sharp3.levels(black_level,gamma,white_level,0,255).coloryuv(autowhite=true)\
.addborders(bord_left, bord_top, bord_right, bord_bot)

result3= sharp3.coloryuv(off_U=blue,off_V=red).levels(0,gamma,255,0,255).addborders(X,0,0,0,$FFFFFF)\
.autolevels().crop(X,0,-0,-0).addborders(bord_left, bord_top, bord_right, bord_bot)

result4= sharp3.coloryuv(off_U=blue,off_V=red).levels(black_level,gamma,white_level,0,255)\
.addborders(bord_left, bord_top, bord_right, bord_bot)

result5= overlay(eval(short),stab_reference,x=est_left,y=est_top).addborders(2,2,2,2,$FFFFFF).Lanczos4Resize(WS,HS)


W2= W+bord_left+bord_right
H2= H+bord_top+bord_bot

short2=short+"S"

source2=Lanczos4Resize(eval(short2),W2,H2)
source3=Lanczos4Resize(eval(short2),W,H)

resultS1= stackhorizontal(subtitle(source2,"original",size=32,align=2)\
,subtitle(result1,"autolevels, autowhite",size=28,align=2))

resultS2= stackhorizontal(subtitle(source2,"original",size=32,align=2)\
,subtitle(result2,"autowhite, manual levels correction",size=28,align=2))

resultS3= stackhorizontal(subtitle(source2,"original",size=32,align=2)\
,subtitle(result3,"autolevels + manual color correction",size=28,align=2))

resultS4= stackhorizontal(subtitle(source2,"original",size=32,align=2)\
,subtitle(result4,"manual colors and levels correction",size=28,align=2))

resultS2H= stackhorizontal(subtitle(source2,"original",size=32,align=2)
,subtitle(result2,"autowhite, manual levels\
 correction",size=28,align=2).histogram(mode="levels"))

resultS3H= stackhorizontal(subtitle(source2,"original",size=32,align=2),subtitle(result3,"autolevels + manual color\ 
correction",size=28,align=2).histogram(mode="levels"))

resultS4H= stackhorizontal(subtitle(source2,"original",size=32,align=2),subtitle(result4,"manual colors and levels\ 
correction",size=28,align=2).histogram(mode="levels"))

result6= stackhorizontal(subtitle(result5,"baseclip for stabiliser -only the B/W clip is used",size=32,align=2)\
,subtitle(stab4,"test stabiliser: dx=horizontal, dy=vertical",size=32,align=5))


Eval(result)

[jagabo]

Experiment: AVSP (use F5 for preview)

[ZQX]

I use the FixEverythingThat'sWrongWithThisVideo() filter. Works perfectly every time.

That would be MC_Spuds, or the latest HybridFUPP!

[2Bdecided]

Some nice work by VideoFred over at Doom9

I'm glad you mentioned that.

It'll get the "I don't use or need AVISynth" crowd suitably jealous.

Though few people have a real need for such processing, it does show what's possible!

Cheers,
David.

[redwudz]

Just a quick word about 'gradation curves'. I ran into a few videos that are flat and low contrast. The several VD contrast/brightness filters had no real effect. So I tried gradation curves. I grab the graph line at about 90 percent and drag it sideways. The same with the bottom of the graph. Now I have contrast and I can easily adjust the brightness and the darkness levels at the same time. Really nice results with some video.

One version: http://members.chello.at/nagiller/vdub/index.html

And if you haven't tried Color Mill, definitely recommended. http://fdump.narod.ru/rgb.htm

Other VD filters: http://www.thedeemon.com/VirtualDubFilters/

[lordsmurf]

Deemon craps itself on my Phenom X4, refuses to run correctly.

[themaster1]

deemon rox (& i got a p4 3g ht)

[redwudz]

Deemon is the filter website, not the program, Daemon Tools, if that's what you mean. For VD filters, there is also: http://neuron2.net/ or MSU: http://compression.ru/video/public_filters.htm

I haven't used Daemon Tools for a while, but I did have problems with it making up a bunch of 'virtual drives' that conflicted with some other programs.

Or do you mean 'Dee Mon Video Enhancer 1.9.2'? http://www.ipmart-forum.com/showthread.php?t=320766 I don't recall if I've tried that.

[lordsmurf]

The last one, Deemon Video Enhancer, which was the site you linked to. It is supposed to use most VirtualDub filters, but unlike Vdub, it does so with multi-core. But for me, it never works, it either goes slower than single-core Vdub, or it just plain ol' craps itself with an error and the app closes.

This was on a Phenom X4, not tried it on my Core 2 Duo yet.

[themaster1]

I frameserve from vdub to the devilish Deemon & it do an awesome job
(i frameserve because neat don't work with it and a couple other filters)
Try that one day with your dual cores...

if it crashes i assume you don't have a valid copy

[Soopafresh]

Avisynth function TemporalDegrain(degrain=3)

Xvid AVI

compare2.avi

[rocky12]

Hey everyone,

I just want to make this thread alive again. Will the thread made by posiondeathray will work? What do I have to download? What do I need to write in my AVS script? This would be for AVS restoration.

[jagabo]

I just want to make this thread alive again.

Why? It's pointless. What you're asking here is something like "I want to learn how to fix cars. Can everyone upload instructions on how to fix cars?" You need to start at the beginning. Take one video you need to fix. Figure out what's wrong with it. Find filters that will fix those problems. Ask for help with particular problems. Nobody has time to write a book on video restoration for your.

Start here: http://avisynth.org/mediawiki/Main_Page

Will the thread made by posiondeathray will work?

Thread? You mean script? Yes.

What do I have to download?

AviSynth, a bunch of filters (note all the LoadPlugin() commands), and the script. Then you can spend a year of your life analyzing the script and what it does.

What do I need to write in my AVS script?

You need a text editor to create and write your script. You need to write commands to do what needs to be done to fix your video.

This would be for AVS restoration.

No, it's for video restoration.

[zoobie]

I remember seeing that goofy script years ago...have fun!

[rocky12]

I can't find AGC, or autolevels, or ImportLimitedSharpenFaster. No, I can't post at Doom9. Myself requested more then 2 or 3 times to be removed or banned. I am not sure if he banned me. I know I been removed from the database.

[manono]

AGC:

http://strony.aster.pl/paviko/hdragc.htm

AutoLevels:

http://www.avisynth.info/?plugin=att...olevels0.3.zip

LimitedSharpenFaster:

http://avisynth.org/mediawiki/LimitedSharpen

No, I can't post at Doom9.

For good reason.

[rocky12]

How does this script look? I get an error in VirtualDub.

DirectShowSource("E:\The Godfather 1 (1972)\video.avi")
Import("C:\Program Files\AviSynth 2.5\plugins\LimitedSupportFaster.avs)
SetMTMode(5)


assumefps(play_speed).trim(trim_begin,0).converttoYV12()
end= source1.framecount()
end2= end-trim_end
frames=end+trim_begin

skip= end2/5
skip0=skip+3
skipend= 3

skipend1= skip0+3

skip2= skipend1+skip
skipend2= skip2+3

skip3= skipend2+skip
skipend3= skip3+3

skip4= skipend3+skip
skipend4= skip4+3

skip5= end2-3

L= trim(source1,0,end2)
LS= trim(source1,0,end2).scriptclip("""subtitle("frame  "+string(trim_begin+current_frame)\
+"  from  "+string(frames)+"   trim_begin-"+string(trim_begin)+"   trim_end-"\
+string(trim_end),x=100,y=60,size=32)""")

sourceT1= trim(source1,0,skipend).scriptclip("""subtitle("frame  "+string(trim_begin+current_frame)\
+"  from  "+string(frames)+"   trim_begin-"+string(trim_begin)+"   trim_end-"+string(trim_end)\
+"  steps: "+string(skip),x=100,y=60,size=32)""")

sourceT2= trim(source1,skip0,skipend1).scriptclip("""subtitle("frame  "+string(trim_begin+skip0+current_frame)\
+"  from  "+string(frames)+"   trim_begin-"+string(trim_begin)+"   trim_end-"+string(trim_end)+"  steps: "\
+string(skip),x=100,y=60,size=32)""")

sourceT3= trim(source1,skip2,skipend2).scriptclip("""subtitle("frame  "+string(trim_begin+skip2+current_frame)\
+"  from  "+string(frames)+"   trim_begin-"+string(trim_begin)+"   trim_end-"+string(trim_end)+"  steps:\
 "+string(skip),x=100,y=60,size=32)""")

sourceT4= trim(source1,skip3,skipend3).scriptclip("""subtitle("frame  "+string(trim_begin+skip3+current_frame)\
+"  from  "+string(frames)+"   trim_begin-"+string(trim_begin)+"   trim_end-"+string(trim_end)+\
"  steps: "+string(skip),x=100,y=60,size=32)""")

sourceT5= trim(source1,skip4,skipend4).scriptclip("""subtitle("frame  "+string(trim_begin+skip4+current_frame)\
+"  from  "+string(frames)+"   trim_begin-"+string(trim_begin)+"   trim_end-"+string(trim_end)+\
"  steps: "+string(skip),x=100,y=60,size=32)""")

sourceT6= trim(source1,skip5,end2).scriptclip("""subtitle("frame  "+string(trim_begin+skip5+current_frame)\
+"  from  "+string(frames)+"   trim_begin-"+string(trim_begin)+"   trim_end-"+string(trim_end)+\
"  steps: "+string(skip),x=100,y=60,size=32)""")

SS= sourceT1+sourceT2+sourceT3+sourceT4+sourceT5+sourceT6


sourceT10= trim(source1,0,skipend)
sourceT20= trim(source1,skip0,skipend1)
sourceT30= trim(source1,skip2,skipend2)
sourceT40= trim(source1,skip3,skipend3)
sourceT50= trim(source1,skip4,skipend4)
sourceT60= trim(source1,skip5,end2)

S= sourceT10+sourceT20+sourceT30+sourceT40+sourceT50+sourceT60


stab_reference= eval(short).crop(est_left,est_top,-est_right,-est_bottom)\
.tweak(cont=est_cont).binarize(threshold=80).greyscale().invert()

mdata=DePanEstimate(stab_reference,trust=1.0,dxmax=maxstab,dymax=maxstab)
stab=DePanStabilize(eval(short),data=mdata,cutoff=0.5,dxmax=maxstab,dymax=maxstab,method=1,mirror=15)
stab2= stab.crop(CLeft,CTop,-CRight,-CBottom).tweak(sat=saturation)
stab3=DePanStabilize(eval(short),data=mdata,cutoff=0.5,dxmax=maxstab,dymax=maxstab,method=1,info=true)

WS= 640(stab)
HS= 352(stab)
stab4= stab3.addborders(10,10,10,10,$B1B1B1).Lanczos4Resize(WS,HS)


vectors= stab2.MVAnalyseMulti(refframes=denoising_frames, pel=2, blksize=block_size, overlap=block_over, idx=1)
denoised= stab2.MVDegrainMulti(vectors, thSAD=denoising_strenght, SadMode=1, idx=1).tweak(sat=saturation)




leveled= denoised.HDRAGC(coef_gain=2.0,max_gain=AGC_max_gain,min_gain=0.5,max_sat=AGC_max_sat,shadows=true)

sharp1=limitedSharpenFaster(leveled,smode=1,strength=LSF_sharp_ness,overshoot=50,\
radius=LSF_radi_us, ss_X=LSF_sub, SS_Y=LSF_sub, dest_x=W,dest_y=H)

sharp2= unsharpmask(sharp1,USM_sharp_ness,USM_radi_us,USM_thres_hold)
sharpX= unsharpmask(sharp2,USM_sharp_ness2,USM_radi_us2,USM_thres_hold2)

sharp3= sharpX.TemporalSoften(temp_radius,temp_luma,temp_chroma,15,2).addgrainC(grain_luma,grain_chroma,0.2,0.2,5)




#backward_vectors = sharp3.MVAnalyse(isb = true,truemotion=true,idx=2)
#forward_vectors = sharp3.MVAnalyse(isb = false,truemotion=true,idx=2)
#frameclip=MVFlowFps(sharp3,backward_vectors, forward_vectors, num=25, den=1, ml=100, idx=2)



result1= sharp3.addborders(X,0,0,0,$FFFFFF).levels(0,gamma,255,0,255).autolevels().coloryuv(autowhite=true)\
.crop(X,0,-0,-0).addborders(bord_left, bord_top, bord_right, bord_bot)

result2= sharp3.levels(black_level,gamma,white_level,0,255).coloryuv(autowhite=true)\
.addborders(bord_left, bord_top, bord_right, bord_bot)

result3= sharp3.coloryuv(off_U=blue,off_V=red).levels(0,gamma,255,0,255).addborders(X,0,0,0,$FFFFFF)\
.autolevels().crop(X,0,-0,-0).addborders(bord_left, bord_top, bord_right, bord_bot)

result4= sharp3.coloryuv(off_U=blue,off_V=red).levels(black_level,gamma,white_level,0,255)\
.addborders(bord_left, bord_top, bord_right, bord_bot)

result5= overlay(eval(short),stab_reference,x=est_left,y=est_top).addborders(2,2,2,2,$FFFFFF)

LanczosResize(640,352)

[manono]

What's the message? You can't just copy and use his script. You have to first understand it and then adapt it to your own video.

What's LimitedSupportFaster? Did you rename LSF, or is that a typo? And you're trying to use this on a downloaded AVI? Why not save yourself the trouble and get the DVD? What you're trying to do is a complete waste of time when there's a better source available. In my opinion.

[jagabo]

rocky12, you obviously have no idea what you're doing. Start with something simple so you can learn something.

[AlanHK]

Read this: Scintilla's Guide to AVISynth Postprocessing Filters

[takearushfan]

Ugh. See, this is why I stopped contributing to MAME. I hate coding. I see those examples and cringe.

[rocky12]

rocky12, you obviously have no idea what you're doing. Start with something simple so you can learn something.

I got a DVD. Alright, but I won't post a clip. How do I write a script?

[AlanHK]

New to AviSynth - start here