How do I know if KaraokeSmith will work on my computer?
It will run on any Windows 10 or 11 (64-bit) PC with at least 8 GB of RAM and about 5 GB of free disk space for the AI models. An NVIDIA GPU speeds things up considerably, but CPU-only mode produces identical results – just slower. The surest way to check: download the free trial and process a song. It runs the full pipeline on up to 10 songs at no cost, so you can confirm everything works on your machine before purchasing.
Does KaraokeSmith require an internet connection?
Almost entirely no. AI separation, transcription, lyric editing, and video export all run locally on your PC. The one exception is optional AI lyric correction, which sends your transcribed lyrics to an Anthropic language model for cleanup – that step requires an internet connection. Everything else works completely offline.
What audio formats are supported?
KaraokeSmith accepts MP3, WAV, FLAC, and M4A files, plus AAC, OGG, and Opus. All format handling goes through bundled ffmpeg – no system codecs or DLLs required.
How long does processing take?
On a machine with a modern NVIDIA GPU, most songs process in 3–8 minutes. On CPU-only machines, expect 20–40 minutes depending on song length and CPU speed. Processing time is a one-time cost – the output is saved and re-editable.
Which GPUs are supported?
KaraokeSmith supports any CUDA-capable NVIDIA GPU (GTX 1000-series and newer). AMD GPU acceleration is not currently supported, but CPU fallback is automatic and produces identical results.
Can I use KaraokeSmith without a GPU?
Yes. If no NVIDIA GPU is detected, all processing automatically falls back to CPU. The output is identical – only the processing time is longer.
What's included in the trial?
The trial includes the complete AI pipeline: vocal separation, Whisper transcription, the full lyric editor, and video export – for up to 10 songs. AI lyric correction is not available in the trial. Purchasing a license removes both restrictions and is valid for unlimited songs.
How does AI lyric correction work?
After Whisper transcription, an Anthropic language model reviews the output for structural errors – wrong homophones, missed words, and repeated phrases that Whisper sometimes produces. It corrects the text only, not the timing, and requires an internet connection. This feature is available in the full version.
Can I edit the lyrics after processing?
Yes. The built-in lyric editor lets you correct transcription errors, insert missing or spoken lyrics, adjust individual word timing, and verify sync against the waveform – all before export. You can also re-edit any previously processed song without re-running the AI.
What is the output video format?
KaraokeSmith exports full HD (1920×1080) MP4 video with synchronized, wipe-style animated lyrics. The output plays on any device that supports MP4 and is compatible with most karaoke players and displays.
Is my audio uploaded anywhere?
Your audio files, stems, and exported video never leave your machine. The one exception is AI lyric correction: if enabled, your transcribed lyrics (text only, no audio) are sent to an Anthropic language model for cleanup. You can use KaraokeSmith without AI lyric correction and nothing is transmitted externally.
Is this a subscription?
No. KaraokeSmith is a one-time purchase of $49.99, activated directly within the app. You own it permanently with no renewal, no monthly fee, and no feature gating after purchase.
What are the system requirements?
Windows 10 or 11 (64-bit), 8 GB RAM minimum (16 GB recommended), and approximately 5 GB of disk space for the AI models. An NVIDIA GPU is recommended for speed but not required.
Editor
Tips & Troubleshooting
Why doesn't the text cursor appear when I click into a field?
This is a known behavior in Chromium-based apps on Windows. The input is active and accepts keystrokes – the cursor just isn't drawn. Clicking another window and clicking back instantly restores it. There's no setting to change; the fix lives upstream in the browser engine KaraokeSmith is built on.
Two words are overlapping and I can't get them in the right order – how do I fix it?
Hold Alt while dragging. By default, the timeline prevents words from overlapping – if you drag a word toward another, it snaps away to avoid a collision. This is usually helpful, but it becomes a problem when Whisper has already placed two words overlapping in the wrong order: dragging either one just pushes it further in the wrong direction with no way to cross. Holding Alt disables the overlap gate entirely, letting you drag a word freely through its neighbour to swap their positions. Alt works the same way in Insert Phrase mode, so you can place a new phrase precisely even when surrounding words are tightly packed. It also means you can intentionally create overlapping words when the song genuinely calls for it.