Supertonic Effects ✨

TTS output rarely lands perfect. There's half a second of silence padding the front, the volume sits quieter than your video's music bed, and the line you just synthesized could do with being a touch faster or deeper. Re-rolling the whole generation for that is the wrong move - trimming and normalizing are deterministic operations, not jobs for a diffusion model. That's exactly what Supertonic Effects is for.

It's quietly the most reusable node in this pack, because it's a standalone AUDIO → AUDIO post-processor that doesn't know or care where the audio came from. Feed it the Supertonic TTS output, a chunk of music, a mic recording - anything that arrives as a standard ComfyUI AUDIO - and it trims, shapes, and spiffs it up before it hits Preview or Save Audio.

What it actually does

Under the hood it's a thin, honest wrapper over librosa, the Python audio-DSP workhorse, and the operations run in a fixed order in the code: trim → clarity → pitch → time-stretch → chorus → normalize. No model, no sampling, no GPU. This is the audio equivalent of unsharp masking - deterministic and instant, the kind of thing the post-processing layer of the ecosystem is supposed to be.

The knobs you'll actually touch:

• trim_silence (on by default) - shaves leading and trailing silence, which is why TTS clips often arrive with dead air at both ends.
• normalize_volume (on by default) - brings loudness to a consistent level so two clips don't fight each other in a mix.
• pitch_semitones - shift the voice up or down, ±12 semitones at half-step resolution. The canonical move: make a narrator sound deeper without re-synthesizing. It only engages at ±0.5 or beyond.
• time_stretch - tempo from 0.5× to 2×; >1 is faster, <1 is slower.
• clarity_boost (off by default) - a preemphasis filter that pushes highs. Useful on muffled audio, prone to adding hiss on clean TTS, so leave it off unless you need it.
• chorus_effect (off by default) - layers a pitch-shifted, 30ms-delayed copy for a richer, thicker sound. Slick on voices, odd on a clean single line.

Everything here is seasoning, not a fix-it machine. The two defaults (trim + normalize) are the ones worth keeping on for every clip; pitch and stretch are the creative moves.

Why it's split out from the TTS node

The pack used to fold these effects directly into the TTS node, and the changelog shows the author pulling them out into this standalone node precisely so they're reusable on any audio source. That's the right call, and it's the right way to structure an audio graph: keep synthesis and cleanup separate so you can swap either stage without touching the other.

Install

Same deal as the whole pack - install via ComfyUI Manager (search "Supertonic"), or:

cd ComfyUI/custom_nodes
git clone https://github.com/Anonymzx/ComfyUI-Supertonic3TTS.git
pip install -r ComfyUI-Supertonic3TTS/requirements.txt

then restart ComfyUI. The node lives under audio/Supertonic.

Where people get burned

The one real trap is librosa. The pack imports it lazily, so plain TTS works fine without it - but this node uses it on every run, and if it's missing you get a clean error telling you to pip install librosa. It's in requirements.txt, so a normal install covers it; the trap is updating the pack without re-installing dependencies.

Second trap: pitch and time-stretch are not model-aware. pitch_semitones resamples the waveform and time_stretch runs a phase vocoder, so push them hard and you'll hear warble and smearing. If you want a faster line, prefer the TTS node's native speed control first and save time_stretch for fine tuning. The README's rule of thumb: effective tempo ≈ speed × time_stretch.