This is an English version of my Japanese article on note, translated with the help of an AI writing tool (Claude) and reviewed by me. I also built the plugin itself largely with AI coding agents.

I updated “Simple Pitch Shift,” an audio filter for OBS Studio that I’ve been developing for a while, to v1.1.0.

obs-simple-pitchshift - OBS Studio plugin

It’s a plugin that processes your mic audio in real time and changes the pitch of your voice. I originally made it to try out OBS plugin development, and the first version prioritized low latency over sound quality.

In this update, I adopted an MIT-licensed audio processing library to improve the sound quality, and made the formant, gain, and processing latency adjustable.

It started as practice for OBS plugin development

I started developing Simple Pitch Shift out of technical curiosity: I wanted to try making an OBS plugin myself.

Pitch shifting is processing that changes only the pitch of audio without changing its duration. If you simply drop or interpolate audio samples, the duration changes too, so it takes some ingenuity.

In real-time pitch shifting, latency is a problem as well as sound quality. For example, when you monitor your own processed voice during a stream, it becomes hard to talk if you hear your voice with a delay.

So in the first version, v0.1.0, I put low latency first. It changed the pitch by moving the read position of a delay line. By crossfading audio taken from two read positions, it achieved pitch shifting with a short latency of about 8.7 ms.

On the other hand, this method had sound quality issues. Changing the pitch could cause wobbling and unnatural components (artifacts), and the release notes at the time stated that I accepted those issues in exchange for low latency.

I had built something that worked for now, but there was still room to improve the sound quality.

Improving sound quality with Signalsmith Stretch

This time I adopted Signalsmith Stretch, a C++ audio processing library.

GitHub - Signalsmith-Audio/signalsmith-stretch: C++ polyphonic pitch/time library (GitHub mirror)

Signalsmith Stretch is a library for time-stretching and pitch-shifting audio, released under the MIT License. Unlike the earlier method of moving the read position of a delay line, it can split audio into fixed time windows and work with its frequency components.

It also has a feature to correct formants, which relate to the character of a voice.

Formants are frequency characteristics produced by resonances in the vocal tract, and they are one of the things that give a voice its character. If you simply change the pitch, these characteristics change too, and the voice can sound very different in character, not just in pitch. By correcting the formants, Signalsmith Stretch can change the pitch while keeping the original voice’s character to some extent.

In v1.0.0, which adopted this library, I set the latency of the audio processing itself to 40 ms. That’s more than the old version’s roughly 8.7 ms, but when I actually tried it in OBS Studio on Windows, I was satisfied with the balance between sound quality and latency.

v1.1.0 lets you choose between sound quality and latency

In v1.1.0, I also made the processing window selectable.

There are three options:

• 20 ms (low latency, low quality)
• 30 ms (slightly low latency, slightly low quality)
• 40 ms (default, prioritizing sound quality)

The processing window is the length of audio handled together for processing. In general, a larger processing window lets you analyze frequencies more finely, but it increases the latency needed for processing by that much. Conversely, a smaller processing window reduces latency but makes the sound quality more likely to degrade.

In this implementation, the latency of the audio processing itself is the same as the processing window: 20 ms, 30 ms, or 40 ms. I actually tried 10 ms too, but the sound quality was quite rough, so I didn’t include it. The quality drop is noticeable even at 20 ms, but I kept it as an option for when you want to prioritize latency.

Formant and gain are adjustable too

In v1.1.0, I added two more sliders besides the processing window.

Formant

You can now adjust the formant in the range from -12.0 to +12.0 semitones. As I explained earlier, formants relate to the character of a voice. While pitch changes how high a sound is, formant can be used to change how a voice resonates and the impression it gives, while keeping the pitch.

Until now, the correction that keeps the original voice’s character when changing the pitch was turned on, but from this version you can adjust the formant itself with a slider.

Gain

You can adjust the gain in the range from -30.0 to +30.0 dB. It’s applied to the audio after pitch shifting.

I’m grateful to be able to use MIT-licensed libraries

It made a big difference that I could adopt an existing library this time instead of implementing the audio processing algorithm myself.

Audio processing, especially pitch shifting, isn’t just a matter of changing the frequency of a sound. It’s a field with a lot to consider, such as sound quality, latency, and the continuity of the audio.

In the first version, implementing a simple algorithm let me try out how OBS audio filters work. On the other hand, improving the sound quality from there would have meant diving into the pitch shifting algorithm itself.

I think it’s really something to be grateful for that I could use an existing library instead of implementing this kind of processing from scratch.

Signalsmith Stretch and the library it depends on, Signalsmith Linear, are both under the MIT License. The MIT License allows commercial use, modification, and redistribution, on conditions such as keeping the copyright notice and the license text. Simple Pitch Shift itself is released under GPLv2, but the MIT License of both libraries is compatible with GPLv2, so they can be used in this setup.

Download

You can download the latest version from Plugin Soko-kun, a free catalog of OBS Studio plugins that I run.

obs-simple-pitchshift - OBS Studio plugin

A prebuilt ZIP is available for Windows x64. Quit OBS Studio, extract the ZIP, and copy the simple-pitchshift folder to %PROGRAMDATA%\obs-studio\plugins. If the plugins folder doesn't exist, you can create it.

To update from v1.0.0, just overwrite it in the same place. Your existing semitone setting is kept, and the newly added settings start at a 40 ms processing window, a formant of 0, and a gain of 0 dB.

If you’re using OBS Studio 32 or earlier, the old-format v0.1.0 is still available.

Release v0.1.0 · AllegroMoltoV/obs-simple-pitchshift

For detailed installation steps and test results, see the README (in Japanese).

obs-simple-pitchshift/README.md at main · AllegroMoltoV/obs-simple-pitchshift

Wrapping up

Simple Pitch Shift started as a project to try out OBS plugin development. The first version prioritized low latency of about 8.7 ms, but in this update I improved the sound quality by adopting Signalsmith Stretch and made it possible to choose the balance between sound quality and latency. With formant and gain adjustments added as well, it’s gradually growing from a simple pitch shift filter into something with more freedom for voice processing. Even something I started out of technical curiosity shows me plenty of things I want to improve once I actually build it.

If you want to change the pitch of your voice in real time in OBS, please give it a try. Bug reports and suggestions are welcome on GitHub Issues.

GitHub - AllegroMoltoV/obs-simple-pitchshift: OBS Studio audio filter plugin to change voice pitch and formant in real time. / 声の高さと声質をリアルタイムに変える OBS Studio の音声フィルターのプラグイン「シンプルピッチシフト」

The original Japanese article: https://note.com/allegromoltov/n/nf5301c73f1d1

Guten Molto! I’m AllegroMoltoV.

This is an English version of my Japanese article on note, translated with the help of an AI writing tool (Claude) and reviewed by me. I also built the plugin itself largely with AI coding agents.

For a while now, I’ve been developing and publishing “Loudness Balance Monitor,” a plugin for OBS Studio. It shows the volume balance between your mic and BGM, the risk of audio clipping, and more in three colors: green, yellow, and red.

obs-loudness-balance-monitor - OBS Studio plugin

I originally built it for my own streams, but I’m grateful that other people have recently started introducing it. In this article, I’ll write about what I was thinking when I built this plugin and about what changed in the new v1.0.0 release.

Introduced by people I didn’t expect

The other day, I found a YouTube video that explains how to install and use Loudness Balance Monitor.

It covers everything from installing the plugin to reading its three indicators: Balance, Mix, and Clip.

I also found it featured as one of the related plugins in a blog post that looks into loudness in OBS in depth.

“How should we handle the loudness of live-stream audio with OBS Studio?” on Semnil’s diary (in Japanese)

OBS Studio を使用したライブ配信音声のラウドネスをどう取り扱ったらよいのか

It’s quite a happy feeling when people I don’t know find something I made because I wanted to use it, and then explain how to use it. I especially appreciate that they explain not just that the plugin exists, but what it can do and how to use it.

Feeling safe to stream matters more than adjusting the volume

I built Loudness Balance Monitor because I wanted to check the balance between my mic and BGM on my own streams.

OBS has a built-in audio mixer, and you can check each source’s volume on its meter. But even if you compare the mic and BGM meters, it’s hard to tell whether the balance is easy on your viewers’ ears.

Also, during a stream it’s hard to hear your own voice objectively, and you can’t keep worrying about whether the volume is right the whole time.

So what I came up with was a way to tell at a glance: “As long as it looks like this, I’m fine for now.” Rather than showing numbers, the plugin takes on the job of telling you “whether it’s okay to keep streaming like this.” The theme of this plugin is to give its users peace of mind as fast as possible.

By the way, loudness is a measure that corresponds to how loud a sound feels to people, and it’s expressed in units such as LUFS. It’s handy for thinking about volume balance, but if the plugin only showed the numbers, users would have to judge for themselves whether they are appropriate. Of course, numbers might be enough for people who know the subject well… but what most streamers want to do is stream, not analyze loudness. That’s why Loudness Balance Monitor shows not only the numbers but also a rating against preset criteria, in color.

Rating three indicators by color

Loudness Balance Monitor watches three indicators:

• Balance: whether there is enough loudness difference between the mic and the BGM
• Mix: whether the combined loudness of the selected mic and BGM is too quiet
• Clip: whether the audio peak level has reached the ceiling

Each indicator is rated on three levels: Perfect (green), Good (yellow), and Bad (red). With the default settings, for example, Balance is Perfect if the mic is at least 6 LU louder than the BGM, Good if it’s at least 3 LU louder, and Bad otherwise. The plugin uses libebur128 to calculate loudness and keeps watching the volume balance and peak levels.

v1.0.0 lets you record and listen back

v1.0.0, the release I’m writing about here, takes this idea one step further.

Until now, the main feature was watching the volume in real time, but I added a new “Recorded monitor.” It records up to 30 seconds of OBS’s final mix and lets you listen to it on the output device you choose. A three-second countdown runs before recording, and playback starts automatically when recording ends.

In the Recorded monitor, you can check both the instantaneous values at the current playback position and the rating for the whole recording.

v1.0.0 also switches the layout depending on the dock’s width and adds a “Usage and terminology guide,” among other changes.

Download

You can download the latest version from Plugin Soko-kun, a free catalog of OBS Studio plugins that I run.

obs-loudness-balance-monitor - OBS Studio plugin

A prebuilt ZIP file is available for Windows x64. The plugin is also implemented and tested in CI for macOS and Linux, but for now you need to build it from source on those systems.

The target version is OBS Studio 33.0.1. Use the recording and playback features with OBS audio set to stereo at 44.1 kHz or 48 kHz. For detailed installation steps and what has been tested, see the README (in Japanese).

obs-loudness-balance-monitor/README.md at main · AllegroMoltoV/obs-loudness-balance-monitor

Wrapping up

Having other people introduce my plugin made me think again about building software for end users. Measuring volume accurately is of course important, but I don’t think users should have to figure out how to interpret those numbers. Rather, I want the software to show a guideline for judging, so that users can focus on what they actually want to do.

The Recorded monitor I added this time is a feature for that purpose too. You check the guideline by color, and if needed, you actually listen to confirm. I hope that helps reduce your worries before a stream, even a little. I plan to keep improving it while using it on my own streams. If you’re having trouble adjusting the volume balance in OBS, please give it a try. Bug reports and suggestions are welcome on GitHub Issues.

GitHub - AllegroMoltoV/obs-loudness-balance-monitor: OBS Studio dock plugin to check the loudness balance between voice and BGM. / 声と BGM のラウドネスのバランスを確かめる OBS Studio のドックのプラグイン「音量バランスモニター」

The original Japanese article: https://note.com/allegromoltov/n/na0467c31f0a0

This is an English version of my Japanese article on note, translated with the help of an AI writing tool (Claude) and reviewed by me. I also built the plugin itself largely with AI coding agents.

I made and released “Simple Mosaic,” an OBS Studio video filter that hides part of the screen with a mosaic.

シンプルモザイク / Simple Mosaic - OBS Studio plugin

Sometimes you want to quickly hide part of your stream with a mosaic. Like the mosaic filter in a drawing app, you just decide the rectangle and the size of the cells. I wanted something that simple, so I built it.

What Simple Mosaic is

Simple Mosaic is a video filter that you add to a source or a group. It has only three kinds of settings:

• Region X and Region Y: the position of the top-left corner of the rectangle to pixelate
• Region Width and Region Height: the size of the rectangle
• Cell Size: the length of one side of a mosaic cell

You set each of them with a slider in pixels. Right after you add the filter, the whole source is pixelated. From there, you narrow the region with the sliders to match the spot you want to hide. Outside the region, the original video is shown as it is.

There are already plugins that can do similar things. For example, a plugin packed with blur features also offers a combination of a mosaic and a rectangular mask. But with that many features, its settings screen is complex too. Simple Mosaic’s selling point is its simplicity: you only decide the region and the cell size.

Download

You can download it from Plugin Soko-kun, a free catalog of OBS Studio plugins that I run.

シンプルモザイク / Simple Mosaic - OBS Studio plugin

A prebuilt ZIP is available for Windows x64. Quit OBS Studio, extract the ZIP, and copy the simple-mosaic folder to %PROGRAMDATA%\obs-studio\plugins. If the plugins folder doesn't exist, you can create it. It supports OBS Studio 33 and later. Rendering is also tested on macOS and Linux in CI, but for now you need to build it from source on those systems.

For detailed installation steps and known behavior, see the README on GitHub.

obs-simple-mosaic/README.md at main · AllegroMoltoV/obs-simple-mosaic

Wrapping up

Simple Mosaic is a filter where you just decide the region and the cell size. If you want to quickly hide part of the screen on your stream, please give it a try. Bug reports and suggestions are welcome on GitHub Issues.

GitHub - AllegroMoltoV/obs-simple-mosaic: OBS Studio video filter plugin to pixelate a rectangular region. / 四角形の範囲をモザイクにする OBS Studio の映像フィルターのプラグイン「シンプルモザイク」

The original Japanese article: https://note.com/allegromoltov/n/n1ab35e9959a6

Guten Molto! I’m AllegroMoltoV.

This is an English version of my Japanese article on note, translated with the help of an AI writing tool (Claude) and reviewed by me. I also built the service itself largely with AI coding agents.

I released “Plugin Soko-kun,” a catalog site where you can look for OBS Studio plugins and list plugins you made yourself. (“Soko” means “warehouse” in Japanese, and “-kun” is a friendly suffix.)

Find plugins for OBS Studio

When a plugin author registers their own GitHub repository, Soko-kun checks the files distributed on GitHub Releases against MalwareBazaar, a database of known malware. If there’s no match, the plugin is listed automatically. Downloads go directly to GitHub, so Soko-kun never stores the files.

You can browse it in a web browser. It supports seven languages, including English and Japanese, and plugin descriptions are automatically translated into the language of the page.

Why I built it

Over the past few months, I’ve built several OBS plugins, such as Simple Pitch Shift, which changes the pitch of your voice, and Loudness Balance Monitor, which checks the balance of your audio levels.

What I learned from building them is that figuring out where to put a plugin and how people will find it is surprisingly hard.

OBS plugins are scattered across many places, such as GitHub repositories and forum posts. For users, it takes some effort both to find the plugin they want and to make sure the file they downloaded is really the one the author published.

There’s one more thing: I build all of my plugins together with Claude Code and Codex. The official OBS forum has a policy of not accepting resources made mostly with AI coding tools. The forum has its own reasons, and that’s fine. But I thought it would be nice to have a separate place where plugins built together with AI can live.

So I decided to make a small catalog where authors register their plugins themselves, the distributed files are checked automatically, and users download directly from GitHub.

Why registration works only from AI agents

Soko-kun accepts plugin registrations only through MCP. MCP is a protocol that lets AI agents use external services. The web pages are for browsing only, and there’s no registration form.

plugin-soko-kun by AllegroMoltoV

I narrowed it down this way on purpose.

The main reason is that I figured the people who would use it are probably developing together with AI agents. If so, the most convenient thing is to be able to ask your agent, right after you finish a plugin, “Register the OBS plugin in this GitHub repository on Soko-kun.” If the plugin doesn’t meet the listing conditions, the agent gets back what’s missing and how to fix it, so you can fix it and register again on the spot.

The other reason is that I didn’t want to complicate authentication. If I also provided a web form, I would have had to build a separate way to sign in to the site. By making MCP the only way in for registration, I only needed GitHub sign-in in one place.

And honestly, I also thought an unusual design would be more fun. A catalog you can register to only from AI agents is a bit odd, and isn’t that fun?

How to use it

As a user, you only need to add the Soko-kun MCP server to your agent.

# Claude Code
claude mcp add --transport http plugin-soko-kun https://soko-kun.allegromoltov.jp/mcp

# Codex
codex mcp add plugin-soko-kun --url https://soko-kun.allegromoltov.jp/mcp

You don’t need to sign in just to search and view details. Ask something like “Find an OBS audio filter plugin on Soko-kun,” and your agent will search for you.

To register your own plugin, sign in with GitHub first: /mcp in Claude Code, or codex mcp login plugin-soko-kun in Codex. Then tell the agent which repository you want to register, and you're done.

The listing conditions are on this page.

Submit a plugin

I just want people to try it first

At the time of writing, the only plugins listed on Soko-kun are the ones I made.

A catalog becomes useful only when plugins from many different people come together, so I want to start by getting someone to use it.

If you make OBS plugins, please try asking your agent to register one. Plugins made with AI and plugins made without it are both welcome. If you’re looking for plugins, I’d be happy just to hear something like “I wish there were a plugin like this” after you search.

Wrapping up

Plugin Soko-kun is a catalog I built on the assumption that both the people who make plugins and the people who use them work with AI agents as their partners. Because I deliberately narrowed down how to register, it will suit some people and not others, but for the people it suits, it should be quite easy.

Bug reports and suggestions are welcome on GitHub Issues.

GitHub - AllegroMoltoV/plugin-soko-kun: Unofficial, MCP-first registry of third-party OBS Studio plugins. / AI と一緒に OBS Studio のプラグインを探したり載せたりできる非公式のレジストリ「配信プラグイン倉庫くん」

The original Japanese article: https://note.com/allegromoltov/n/n694bafe17a02

Guten Molto! I’m AllegroMoltoV.

This is an English version of my Japanese article on note, translated with the help of an AI writing tool (Claude) and reviewed by me. I also built the plugin itself largely with AI coding agents.

I made and released “AI Assistant,” an OBS Studio plugin that lets you talk to an AI about streaming problems in a chat inside an OBS dock.

ai-assistant-for-obs - OBS Studio plugin

What AI Assistant is

AI Assistant is a chat that lives in an OBS dock. When you type what’s bothering you into the input box and send it, the AI reads information like the following to narrow down the cause:

• The OBS state and statistics (resolution, FPS, dropped frames, and so on)
• The list of scenes and sources, and the settings of each
• The last 300 lines of the OBS log

If something still isn’t clear after reading all that, the AI asks you one or two questions at a time. For things that are better looked up on the web, such as known bugs, it searches and adds links to its sources in the answer.

So that you can start even if you can’t put the problem into words, an empty conversation shows seven buttons with common problems, such as “The screen is black” and “My microphone is not heard.” Just press one to start a consultation.

Answers are displayed as they arrive. Even for long answers, the first words appear within a few seconds.

Background

When you run into trouble in OBS, such as “nothing shows up,” “no sound,” or “the video stutters,” it’s surprisingly hard to put the cause into words yourself. If you don’t know which settings are involved, you can’t even come up with the words to search for.

I had been thinking it would be handy to have someone to consult who does the whole routine for you, like Amazon Q Developer does: look at the state of the environment, investigate, and fix the settings if needed. I figured that if I could put that inside OBS, I could ask about streaming problems in the same way, so I tried building it.

Approval before every change

What I was most careful about in letting an AI touch OBS was not leaving everything to the AI.

When AI Assistant wants to change an OBS setting or add a filter or source, it first shows the change as a proposal. The proposal shows the item to be changed and its values before and after, for example, “Mic Audio Monitoring: Monitoring Enabled → Monitor Off.” The change is applied only when you press “Approve and apply,” and nothing changes if you press “Reject.”

What it can and cannot do is clearly separated.

What it can do

• Investigate and explain the cause
• Change settings inside OBS
• Add and edit noise suppression, gain, compressor, and crop/pad filters
• Add and edit application audio capture, browser, and color sources
• Delete things that the AI itself added

What it cannot do

• Delete scenes, sources, or filters that a person already created
• Start or stop streaming and recording
• Operate anything outside OBS, such as Windows settings

It can only delete things it added itself. It marks what it adds, and the plugin refuses to delete anything without that mark. If you ask it to delete something a person created, the AI suggests hiding, disabling, or muting it instead.

Cost and the data that is sent

The plugin is free. However, the AI uses the OpenAI API, so usage fees are billed to your own OpenAI account. Before you start, issue an OpenAI API key and save it in the dock’s “Settings.” (The key is stored encrypted with Windows DPAPI.)

The model is gpt-6.1-terra (effort: low). In my own checks, one consultation cost about 0.012 USD on average and about 0.055 USD at most. The most expensive one used web search three times, and most of the cost came from the web searches.

To answer, the plugin sends your messages and the OBS state, settings, and log that the AI read to OpenAI. The settings include scene and source names, text source contents, file paths, and browser source URLs. Images and video are not sent.

• Stream keys, keys in URLs, passwords, and tokens are masked before sending
• Before sending, the content is checked with the OpenAI moderation API, and content that may violate the usage policies is not sent
• Conversation history is not saved to files

Download

You can download AI Assistant from Plugin Soko-kun, a free catalog of OBS Studio plugins that I run.

ai-assistant-for-obs - OBS Studio plugin

A prebuilt ZIP is available for Windows x64. Quit OBS Studio, then copy the ai-assistant-for-obs folder in the ZIP to %PROGRAMDATA%\obs-studio\plugins. If the plugins folder doesn’t exist, you can create it with that name. After starting OBS, show “AI Assistant” from the Docks menu.

I tested it on OBS Studio 33.0.1. It can’t be used on macOS or Linux, because saving the API key is supported only on Windows.

The README explains how to use it in detail.

ai-assistant-for-obs/README.md at main · AllegroMoltoV/ai-assistant-for-obs

Wrapping up

I introduced “AI Assistant,” a plugin that lets you consult an AI about streaming problems inside OBS.

Letting an AI touch OBS is convenient, but it’s also scary that settings might change without you noticing. So I built it so that you always check the values before and after with your own eyes before approving.

If you’ve ever been stuck with a streaming problem without knowing what to look into, please give it a try.

Bug reports and suggestions are welcome on GitHub Issues.

GitHub - AllegroMoltoV/ai-assistant-for-obs: OBS Studio dock plugin to troubleshoot streaming problems with an AI assistant. Uses your own OpenAI API key. / 配信のトラブルを AI に相談できる OBS Studio のドックのプラグイン「AIアシスタント」

The original Japanese article: https://note.com/allegromoltov/n/nd43e81bb2fa2