A single native binary, no Python runtime, no shell scripts, no mismatched libraries to chase.
Just download one file and run.
It’s a full-featured editor, with a powerful AI generation pipeline built right in (2D & VR).
So you can generate, edit, refine. Or just generate and play. Or even, just play.
The editor
Timeline : frame-accurate point editing, rich selection (peaks/valleys, top/middle/bottom, by-chapter, left/right of playhead), transforms (equalize, isolate, ripple-delete, repeat-stroke, snap-to-playhead magnet, etc.), parametric stroke generation, all with live preview, and a plugin / macro system for repeatable edits.
Multi-axis : full multi-axis scripting (stroke, surge, sway, twist, roll, pitch). Open several scripts/axes as lanes side by side and ghost one over another to compare and reference while you edit.
Funscript Doctor : scores your script’s quality, flags issues (speed limits, gaps, missing strokes…), and offers one-click auto-fixes.
Chapters, bookmarks & heatmap : chapter bar with custom types, bookmarks, a colored heatmap bar.
Unlimited undo/redo : with a visual history browser, that even survives closing and reopening a project.
AI generation : try it free, never forced to buy
Turn raw video into a draftable script in one pass, then refine by hand :
FunGenerate : proposes the next 20 seconds of funscript from the playhead, shown as a live ghost over your script; accept or reject before it commits. So you can judge the AI’s quality on your content before deciding anything.
Full auto-generation : run the AI + tracking pipeline over the whole video.
Live recording mode : drive the motion yourself (mouse, controller) while it plays and record it live.
Quick-generate from Welcome : drop a video in and go.
Two tiers, no lock-out:
Free : base algorithmic generation for everyone, plus the AI generation itself behind a short metered wait. You can genuinely try it.
Pro : removes the wait and unlocks the dedicated high-quality 2D and VR AI models.
Playback & devices
Built-in player : hardware-accelerated decoding; scrub and preview without a separate player.
Device output : drives single and multi-axis hardware over Buttplug, T-Code and Handy, with an on-screen position gauge.
Cross-platform, properly
One binary per platform (Windows / macOS / Linux), localized in multiple languages.
That is really exciting work you’ve done during the last year. Congratulations on reaching this level of development and thank you and your contributors for this amazing contribution to this small community.
I can’t wait to try it out, once I can carve out the time, especially with two very reliable voices having been actively nothing short of praising the new FunGen to the skies. And reading the description I can see why.
I do wonder about a few things though:
Is it possible to place the actions more precisely? This is usually needed for low fps videos while making audio-based scripts. Is there a music/beat-based GUI mode?
Is or will there be a general plugin system as in OFS, where it’s possible to make lua (or a different language ) integrations for various purposes?
Furthermore considering AI: Is it possible to train and use an own model?
I guess we are missing a recent DirectML.dll with the right entry point on this older version of Windows. Will see if this can be solved, should be an additional 15MB to add in the release though.
Thanks randomus, really appreciate the kind words.
Precision for low-fps / audio-based scripting
There’s a beat/music workflow built in, but right now, by design, its snaps to frames.
Could revise that if necessary.
Plugin / macro system
There’s a set of built-in transforms ( around 20: amplify, speed-limit, RDP/Savitzky-Golay smoothing, autotune, resample, etc.) plus one-click quickfixes, and a macro recorder.
What there isn’t yet is an open user-scriptable runtime like OFS’s Lua, where you write arbitrary scripts.
Your own AI model
No, doesn’t work like this, at least for now.
Currently considering options for the model to learn from user corrections.
Devices : yes, direct
Exactly
The player is built in and that same playback drives single and multi-axis hardware directly over Buttplug, T-Code and Handy.
No second app in the loop, load the video, hit play, it streams to the device.*
You can use FunGen for free without the Pro features. Try it first to see whether it meets your requirements, and then decide whether you’d like to support its further development.
Yes, but - how much is it? It’s a fair thing to ask. I don’t see a price, or a payment model.
I already have a free workflow with OFS and multiple methods to get it to generate something that I can manually tweak into something good - I’m assuming Fungen also needs manual tweaking. It might do a better job, who knows? But before I come to rely on features that might be locked behind a subscription now or in future, I’d like to know the terms of the deal. That’s a reasonable request for any potential purchase.
Imo - and I haven’t tried out your implementation yet, so grain of salt - it would make sense to have actions generated to a rhythm, not frames, when using a music workflow/GUI mode. In OFS’ tempo mode I can shift the offset in steps of 1 ms, and the actions can be placed exactly to notes and bars, rather than video frames. While 1ms is possibly overkill, certainly it needs a better resolution than the video framerate. This would also allow for uniform action lengths and with it speeds within uniform patterns , which - depending on the transforms - can generate unclean sequences because of the difference of fps and bpm.
Tl;Dr: For music, we need a better resolution than the video framerate.
Looking forward to see what the macro recorder and the transforms are all about. But is a full plugin-system (e.g. OFS lua extensions) intended to be implemented? I kinda heavily rely on some of my extensions and have a few ideas for Multi-Axis scripting.
I will get back to you once I’ve dabbled around with it a bit and maybe you’ll want to implement parts of it anyway like the other transforms.
There you can also see that a few free AI previews are available each day and that there is a “perpetual” option that looks like a one-time payment model.
The tool itself is free; there are no trials as they aren’t needed. There is a Pro version of it that you can upgrade to however.
I believe you can use the free AI functions with no limits, and you get 3 daily 20-second trials of the Pro AI
Same. I looked into this, but I can’t find anything related to AI scripting unless I purchase the “Pro” tier. That’s unfortunate… before I purchase anything, I need to evaluate how effective the new AI functions are (and in particular if they’ve improved since the last version.) Sorry!
So I registered for 99$ lifetime so I guess its better to get in early. Will be unpayable if AI scripting gets perfect. I guess >500$
Here is the hard truth about Fungen AI scripting quality atm:
Turning points close to 1 frame precise. Endpoints not perfect but OK. Some times 2 seldomly 3 off.
movement length is close
no unnecessary points set
oscillating in center no true heights set or recognized.
full in / full out points unknown to tracker
no speed changes scripted slow out first 1/3 fast out last 2/3, no slowdown by turn
not all movements recognized, some small gaps
detection problems after scene and cam changes
for perfect scripting still every single point has to be edited
Overall script quality on scale 2/10 “moves the device resembling to what happens on screen” Perfect script time gain to complete - minimal. An untouched AI script should also been sent to an event on ES. It will for sure be last places but it can be seen how it will climb in places reached, or not. Editor: 8/10 and better than OFS
The editor is free and will quickly replace OFS. No need for Iframe slave file and scripting in 1080p blurryness. Most tools like the quickfix ones are integrated and easy to use with previews. If a feature is good you will get it, fast response from @k00gar. Bugs are squashed fast within days at least for now. He is listening closely to the discord community.
So should you buy Fungen anually or lifetime when I show AI generation quality is way back compared to an award winning script or the winning script in an event? Yes as it will get there as autonomous car driving will get there one day.
Is it helpful to a scripter atm? Absolutely not, its zero time gain atm, it is like rescripting a bad out of sync script. Its a full redo. But seeing what the cat (Koogar) has done to the editor in weeks is amazing and I have zero doubt there will be soon scripts made by Fungen AI that will reduce scripting time and at the end replace it fully with minimal preference editing. This will take time, months maybe years but he will get there. So its a leap of faith but I really think it is worth supporting the project. And if you dont want to support him, then just use the editor without AI that is by far the best out there, always free he said.
) integrations for various purposes?
.
