HeadMania

I have played with many music players throughout the years—foobar2000, JRiver, Roon, Audirvana, and HQPlayer. All of them have their strengths, and I have genuinely enjoyed using them. I am deeply grateful that they exist and for the hard work their developers have put in, as they have consistently improved my listening experience over time.

My journey into digital upsampling began years ago when a friend told me about his experience listening to a Chord DAC with and without the M-Scaler attached. That got me curious enough to look into software upsampling. I started testing the upsampling engines in Roon, Audirvana, and downloaded the trial version of HQPlayer.

I immediately noticed the differences in sound quality. Across the board, the sound became more natural, holographic, more textured, and significantly cleaner. Of course, the results varied depending on the engine. I liked the r8brain implementation in Roon, but I actually found the upsampling results from Audirvana to be superior. Ultimately, because Audirvana acted as a proper aggregator for the streaming services I use (like Qobuz and Tidal), I chose to stick with it—even though HQPlayer sounded arguably better overall, albeit much more expensive.

Recently, my friend Sandu from Soundnews mentioned upsampling again, and it got me thinking. I realized that today, we have easy access to staggering amounts of processing power. A modern multi-core PC should be able to handle extreme mathematical upsampling effortlessly, completely removing the processing burden from the tiny, underpowered silicon chips inside our hardware DACs. I realized it was time to finally build something that combined all the features I had ever wanted in all the players I had used over the years.

Long story short: I built the HeadMania.org UpSampler.

Video presentation here:

Today, after countless iterations, I can confidently say that I prefer the sound quality of the HeadMania UpSampler over both Roon and Audirvana, and I dare to say it rivals HQPlayer. But beyond sheer sound quality, it was built to do things that no other software on the market does.

It is designed to be the ultimate digital transport, complementing existing ecosystems with the following core capabilities:

• System-Wide Windows Capture (An Industry First): No other audiophile player does this. You can capture audio from any app natively (Spotify, YouTube, Deezer, games) and upsample it to 1-Million taps before sending it to your local DAC or casting it to a network streamer. This effectively bypasses the notoriously bad USB inputs that many DACs have.
• Ultra-Low Latency Adaptation: Custom apodizing filters engineered specifically for zero lip-sync delay in movies and gaming.
• Local File Playback: Bit-perfect rendering of local FLAC, MP3, and WAV libraries.
• Headless Network Upsampler: Completely independent of Windows processing. It runs a local server with a clean, responsive Web UI that you can access from any phone or tablet on your network to configure settings and outputs.
• UPnP-to-UPnP DSP Bridge (Completely Unique): Some audiophile software can accept a UPnP stream as input. But no other software on the market takes that incoming UPnP stream, runs it through a massive 1-Million tap FIR processing engine, and then re-casts the perfected signal back out to your hardware DAC over standard UPnP. HeadMania does. You can use any app on your phone (like mConnect or BubbleUPnP) to stream to HeadMania, which upsamples and processes the signal, then seamlessly forwards it to your network DAC. It essentially harnesses the raw processing power of your PC to massively upgrade the performance of your DAC, no matter where it sits on your network. And because this headless pipeline completely bypasses the Windows audio kernel, the signal remains mathematically pure—meaning the sound quality is identical to what you would get running a custom Linux or macOS core.
• Hardware Synergy Engine: Automatically tailors the digital output stage specifically for your physical DAC topology (Discrete R2R, Pure 1-Bit Native DSD, or Commercial Delta-Sigma) with optimized quantization and true-peak modulator ceilings.
• Kinetic Transient Engine: Shatters the “long-filter transient smearing” paradox. It actively monitors the audio envelope and dynamically shifts convolution weighting to a lightning-fast cascade filter the exact microsecond a drum impact occurs. You get explosive slam combined with the immense 1-Million tap immersive, textured and natural sound experience.

Here is a deeper look at what makes it special.

What Makes It Special?

1. The 1-Million Tap “Extreme” & “Dark Extreme” Filters
Most hardware DACs use around 256 taps for their digital reconstruction filters. The HeadMania UpSampler’s Extreme profile uses over 1,000,000 taps. When I built the first implementation of this filter (based purely on linear phase), I immediately noticed beautiful improvements in holography, natural tonality, and transparency. The imaging became completely decoupled from the headphone drivers.

However, when comparing it to the basic r8brain upsampler (running in minimum phase), I felt the pure linear phase filter lost a bit of transient attack and tactility. I wanted both.

So, I built the Extreme Hybrid filter. It applies an apodizing correction that keeps all the magical transparency and holography of the 1M-tap linear phase, but injects the visceral tactility and punch back into the transients. It synergizes phenomenally with DACs like the Rockna Wavedream Signature (which also utilizes a hybrid filter architecture). When you listen through a headphone like the T+A Solitaire P, the presentation becomes:

holographic
tactile
mesmerizing
lifelike

I also recently added the Dark Extreme and Dark Extreme Hybrid profiles. These maintain the massive 1M-tap processing but apply a strict 0.85x Nyquist cutoff for a smooth, organic treble rolloff. If you have an aggressively bright headphone (like an HD800) or a clinically analytical DAC, this filter injects an undeniable vintage warmth into the chain without sacrificing macro-details.

2. The Cascade Filters & The “Explosive Extreme” Engine
The story behind our Cascade filters is actually a lesson in humility. After finishing the 1-Million tap Extreme Hybrid filter, I was incredibly excited to test it on the Linn Selekt Edition Hub with the dual-mono Organik DAC. I expected it to sound spectacular, just as it did on the Rockna. To my surprise, the sound didn’t improve. In fact, it became a bit disjointed and harsh on the transient leading edges.

I dug into what Linn was actually doing under the hood. The Organik DAC isn’t a traditional R2R or Delta-Sigma chip; it uses an ultra-fast FPGA feeding a discrete Pulse Width Modulation (PWM) stage. I realized that by feeding a massive, mathematically dense 1-Million tap signal into a PWM architecture, I was actually forcing the Linn’s internal FPGA to work incredibly hard to deconstruct the signal to modulate the pulse widths. The two heavy DSP engines were fighting each other.

So, I built the Cascade and Cascade Hybrid filters. Rather than trying to overpower the DAC, these filters provide a clean, ultra-short impulse response. They apply the vital apodizing phase correction to prevent pre-ringing, but they deliberately step out of the way and defer the heavy reconstruction lifting to the DAC’s internal FPGA. It is the perfect handshake between software and hardware.

But I didn’t stop there. Having both the massive 1-Million tap filter and the lightning-fast Cascade filter in my arsenal sparked a new idea. What if we didn’t have to choose?

With today’s release, we introduced the Kinetic Transient Engine, powering the new Explosive Extreme profile. This is a groundbreaking dual-engine architecture designed to literally shatter the Fourier Uncertainty Principle. It runs the pristine 1-Million tap Extreme filter in parallel with the lightning-fast Cascade filter. The engine actively monitors the mathematical derivative of the audio envelope. The exact microsecond a drum impact occurs, it dynamically shifts the convolution weighting toward the fast Cascade filter. Once the transient energy dissipates, it seamlessly restores the full 1-million tap density.

You get the explosive, un-smeared transient slam of a short filter, combined with the 240dB stopband rejection and immense soundstage of a 1-million tap filter during sustained notes.

3. Live Streaming Intercept & UPnP Bridging
You aren’t restricted to local FLAC files. By clicking “Stream Live,” the engine captures the audio from any app playing on Windows—like Deezer, Spotify, or YouTube (just make sure that app isn’t used in exclusive audio mode). Even though the input has passed through the Windows mixer, we drastically improve the signal with 64-bit processing, 1M-tap upsampling, and True Peak Limiting.

But here is the killer feature that no one else does: you can route this improved stream to your local USB DAC, OR wirelessly to a UPnP endpoint. If you love an app like Deezer that doesn’t natively support UPnP streaming to high-end gear, HeadMania acts as the bridge. You play Deezer locally, we intercept it, upsample it, and cast it to your high-end streamer over the network. Because of the heavy 1-Million tap processing and the network buffering, this isn’t strictly “real-time” (there is a noticeable delay when you press play/pause), but the resulting sound quality is pristine and bit-perfect.

4. Hardware Synergy Engine
Not all DACs handle quantization noise the same way. The Hardware Synergy Engine automatically tailors the final 64-bit to 24-bit output stage specifically for your physical DAC topology. If you run a Discrete R2R ladder DAC like my Rockna, it completely bypasses dithering to eliminate LSB-switching thermal grain. If you use a Pure 1-Bit DSD engine, it enforces a strict -3.5 dBFS True Peak ceiling to prevent modulator instability, combined with Flat TPDF dither. And for standard Commercial Silicon (like ESS or AKM chips), it uses highly optimized 3rd-order noise shaping. This guarantees that the digital output is mathematically optimized for the exact silicon architecture receiving it.

5. Apodizing “Live Hybrid” Filters
Linear phase filters sound incredible, but they introduce ~740ms of latency (group delay). For movies or gaming, we built the Live Hybrid filters which drop the latency down to an imperceptible 11 milliseconds for perfect lip-sync, while still deleting pre-ringing.

6. The Network Renderer (Bypassing Windows) (Completely Unique)
If you want the absolute highest sound quality possible, HeadMania can operate as a headless UPnP server. When you play local files or whatever application in this mode, the audio is routed directly from the file to the HeadMania DSP engine, and then cast directly to your DAC via the network. Windows audio processing is completely bypassed. The Windows kernel never touches the signal. Because of this, it does not matter that the software runs on Windows—in this mode, the bit-perfect purity is identical to what you would get running a custom Linux kernel or a Mac.

7. Psychoacoustic DSP & Dynamics
I wanted the ability to tune my headphones to the room or my mood without destroying the signal integrity. The DSP engine includes:

• AutoEQ Integration: Instantly flatten the frequency response of hundreds of headphones, automatically applying correction curves aligned to the industry-standard Harman 2018 acoustic target for a naturally balanced sound out of the box.
• Dynamic Bass Thump: This isn’t a shelf EQ that muddies the mids. It’s a transient enhancer. When a kick drum hits, you don’t just hear the drum—you feel the drum.
• Speed/Transients & Harmonic Exciter: Sharpen the leading edge of notes for faster PRaT, or add a subtle analog-style shimmer to the top end.
• Audiophile Crossfeed: A high-quality Bauer/Chu Moy hardware simulation that reduces extreme stereo separation to emulate a natural speaker presentation. While I personally keep this off to preserve absolute transient clarity, many users love it for reducing headphone fatigue on hard-panned stereo mixes.
• Parallel Detail Enhancer: Upward compression that brings out micro-details (like hall reverb or breath) without crushing the dynamic peaks.
• Soundstage (Width): Mid/Side processing that scales up the Side channel to widen the stereo image without altering the center phantom image (where the vocals sit).
• Sound Profiles (Action / Dialogue / Night): One-click presets designed for movies and gaming, allowing you to boost dialogue intelligibility or compress dynamics so you can watch movies at night without waking the family.

8. True Peak Protection & O(1) Limiting
Pushing heavy EQ or adding width inherently creates inter-sample peaks that will clip your DAC’s output stage. Many digital players use cheap, feedback-clamp limiters that squash the life out of the music. To fix this, I completely rewrote the protection stage into an O(1) monotonic sliding-window feed-forward True Peak limiter. It uses a 10ms lookahead to perfectly anticipate peaks, applying smooth gain reduction without ever hard-clipping the signal. The music retains its dynamic range, the transients still hit hard, and the distortion stays at absolute zero.

System Requirements

This isn’t a weakness; it’s the reality of extreme mathematics. The 1-Million tap Extreme filters require a modern CPU. If you run the Extreme filters on an old laptop, it will probably stutter. However, the lighter filters (like Natural, Linear, or Cascade) run effortlessly on older machines. If you have the hardware to run the Extreme profiles, it brings the performance of your existing DAC up to another level entirely.

This was built to be the ultimate digital transport for the HeadMania community. Try the Extreme Hybrid filter, but make sure to experiment with the other profiles (like Cascade or Linear) to find the perfect synergy with your specific DAC and its internal filtering.

Once you find that perfect match, something magical happens. The truth is, ever since I finalized the Network Renderer function, I simply cannot stop listening to music across all my systems—whether it’s the headphones and Martin Logans on the Rockna Wavedream Reference Signature, or the Piega Coax 611 speakers through the Linn Selekt Dual Mono Organik. I just can’t get enough, and I constantly find myself listening deep into the night.

Queue up your favorite song and just listen. It hits hard, it goes deep, and it might just wipe the floor with you.

---

A Quick Disclaimer & A Note on “Free” Software

Please keep in mind that HeadMania is a one-man passion project, built entirely in my spare time to solve my own audiophile needs. It has been extensively tested on Windows 11 against my specific use-cases, but it is still early days, so you might encounter a quirk on different setups. I appreciate your patience as the software matures.

Secondly, while the core version of this software is free, please do not mistake “free” for “amateur.” I know many of you are sitting in front of incredibly expensive systems, looking at the word “free” and immediately assuming it can’t compete. Don’t make me do it—I’ll start charging you a percentage of your system’s worth. 🙂

I am releasing this freely out of love for the audiophile community, and I also published a Patreon page for those who want to show their support for its ongoing development, which will be greatly appreciated.

However, from an engineering standpoint, this engine is built on a lifetime of software architecture. I am a Computer Science engineer with over 20 years in the field. From winning national and international software contests in my youth, to serving as a University Teaching Assistant, to my current role as a Senior Director at a global corporation—building high-performance architecture is my life’s work. This project simply represents the intersection of my two greatest passions: extreme software engineering and high-end audio.
You can download the UpSampler from here.

Edit.
The last version contains a complete port to MacOS (Apple Silicone). Get it while it’s free. The mac will at one point move under patreon.

Also, something very important that is also mentioned in the below video. In the context where you want to use it to capture the sound from windows, you
1. Select a DIFFERENT default sound output than the you want to use as output. This input should stick to lower bitrate and sample rate(44khz, 48khz). Keep similar to the below. If you only have one output check out step 3.

2. IF you select from headmania a local dac output instead of upnp output, again, it’s very important that Windows HAS ANOTHER default input with low bitrate. We don’t want windows to apply it’s lower quality upsampling. Leave the upsampling to the HeadMania upsampler.
Also, it’s important that for the device you select as output (from HeadMania player, but is NOT the default output in windows) you have the 24 bit, 384000hz selected, so the dac can receive the upsampled stream.

3. If you only have one output in windows available, the easiest way to fix :

The Solution: VB-Audio Virtual Cable
VB-Cable is a free, lightweight driver that creates a bit-perfect virtual “Playback” device in Windows.

Step 1: Install VB-Cable

Go to vb-audio.com/Cable and download the Windows version.
Extract the ZIP file.
Right-click VBCABLE_Setup_x64.exe and select Run as Administrator to install it.
Restart the PC.
Step 2: Configure Windows (with the step 1 example and settings)

Click the Speaker icon on the Windows taskbar.
Select CABLE Input (VB-Audio Virtual Cable) as the system’s default output device. (Now, all PC sound, Tidal, YouTube, etc., is playing silently into this virtual cable).

4. If you are using Tidal, Qobuz or any other players from which you want to upsample through headmania, it’s very important that you disable the exclusive output and leave system default in the output settings of that Tidal, Qobuz or whatever else player.

Brief how to use video. Please watch: