Merge branch 'master' of github.com:kanjieater/AudiobookTextSync

ccos89 · Mar 20, 2023 · 465b167 · 465b167
2 parents 0d6f6b4 + 4844b40
commit 465b167
Showing 1 changed file with 3 additions and 1 deletion.
diff --git a/readme.md b/readme.md
@@ -63,6 +63,8 @@ You need two copies of your file. One in "<full folder path>" and one in `<full
 
 At this point I would recommend reading from the texthooker instead of a sub. (CTRL+SHIFT+RIGHT in mpv to set offset as the next sub). Then you can see the next line coming in the texthooker, and not be distracted by subtitle jumps.
 
+Update: The timing is much more accurate, but it still makes sense to show what going wrong could look like
+
 https://user-images.githubusercontent.com/32607317/219973663-7fcac162-b162-4a02-839c-0be2385f6166.mp4
 
 
@@ -114,7 +116,7 @@ curl --header "Content-Type: application/json" \
 # Troubleshooting
 You might see various issues while trying this out in the early state. Here are some of the pieces at work in sequence:
 ## Stages
-1. (not pushed yet) Filter down audio to improve future results - slow & probably not heavy cpu or gpu usage. Heavier on cpu
+1. Filter down audio to improve future results - slow & probably not heavy cpu or gpu usage. Heavier on cpu
 2. split_run & stable-ts: Starts off heavy on CPU & RAM to identify the audio spectrum
 3. stable-ts: GPU heavy & requires lots of vRAM depending on the model. This is the part with the long taskbar, where it tries to transcribe a text from the audio. Currently the default is [tiny](https://github.com/openai/whisper#available-models-and-languages). Ironically tiny, does a better job of keeping the phrases short, at the cost of accuracy of transcription, which since we are matching a script, doesn't matter. Also it runs 32x faster than large.
 4. Merge vtt's for split subs