PyResampler

Python GUI toolkit for passing down Resampler arguments (single and batch resampling)

---

To do:

• UI improvements
• Control output speed
• Use wine for resamplers (for mac and linux)
• Switch to pure python based FFpmeg and FFprobe (so mac and linux can use them too)
• Use custom fonts
• Multi-threaded resampling
• Option to generate frq files with Harvest
• Accept any type of audio files
• Convert to wav (temp file) before sending it to the resampler
• Fader2.exe support
• Audio length/duration editor
• Render long audios in chunks
• Diffsinger vocoder inferencing (CPU only atm)

---

How to use:

• Install necessary modules in order for the GUI script to run.

  pip install -r requirements.txt
  

• The GUI will save the resampler and output directory in the config.yaml file.
• You also need to download ffmpeg, ffprobe, and ffplay in order to for this toolkit to work (windows build is already included in the assets folder)
  • https://www.ffmpeg.org/download.html#build
• After downloading, place the files in Assets/Ffmpeg/(your os [win, macos, linux])
• For macos and linux devices, you need to have Wine binaries so you can run the exe resamplers

---

Main UI

Audio input & Resampler directory

• You can input audio via drag and drop or via Add files button

UI Tabs

• Single Resampler picker - single resampler rendering

• Batch Resampler picker - multiple resampler rendering

• Audio trimmer/duration editor
  • Left/Right lines: Similar to left blank and cutoff, you can drag them to cut the audio and only render the selected part
  • Red line: Playhead, click the waveform to play on that exact position

• Settings
  • Themes: light, dark, and system theme
  • Wine path: click the Detect button to automatically find the wine path (usable for MacOS and Linux)
  • Open button: if the wine path detector fails, you can click this to manually find the wine path binaries
  • Open Output: Opens the output folder after rendering the input files
  

Processing Parameters

• Flags: Global resampler flag input, (applies the flags to all input files, flags varies for each resampler)
• Threads: Resampler thread and Harvest frq thread, the more threads to set, the faster the execution of the processes
• Clear Temp Cache: Clears the temp files inside the cache folder
• Pitch: The base pitch of the output render, you can input herts or the musical notation of the pitch (440, 390, 20, C4, G#3, D5)
• Speed: Speed of the rendered audio (slow or fast)
• Follow input pitch: Detects the average pitch of the input audio/s so you don't have to guess them, you can do the other method by making the pitch entry blank so the resampler skips the pitch input (*note: some resamplers will error or the output is unexpected from the result if the pitch entry is blank)
• Volume: Volume of the rendered audio
• Modulation: Modulation parameter of resamplers, uses the detected f0 of the audio so it copies the actual pitch of the audio (0 = flat, 100 = actual pitch, 200 = exaggerated, -100 = inverted pitch)
• Spectrogram: Generates the image spectrogram of the rendered audio
• Harvest.frq: Uses Harvest algorithm to generate the frq files instead of the resamplers (very fast and accurate, very useable if the resampler also uses frq files for the f0)
• Only frq: Outputs only the frq files (uses harvest)