A locally running AI stem separator developed by a team registered in Jiangsu, China, with the domain registered in November 2024. It imports audio or video files and splits mixed tracks into independent stems across three scenarios: music (vocals, harmony, accompaniment, or individual drums, bass, guitar, piano), speech (removing ambient noise to isolate clean dialogue), and video (separating dialogue, music, and sound effects from film or video). Processing is entirely local, so files are never uploaded to a cloud, which is the product’s core privacy claim. It supports GPU acceleration but only on NVIDIA cards from the M250 upward; AMD is unsupported, and machines without a compatible GPU fall back to slower CPU processing. Batch processing handles thousands of files at once. Input formats cover MP4/MKV/FLV/AVI/WEBM for video and MP3/WAV/FLAC/AAC/M4A for audio, with WAV/FLAC/MP3 output. Pricing has a permanently free Basic tier that includes standard models, batch processing, GPU acceleration, and WAV/FLAC output, plus a Pro tier at $19/month (listed as marked down from $29, a 35% reduction in January 2026) that unlocks top-tier vocal, accompaniment, and film audio models, and does not auto-renew. RightAIChoice has re-verified the listing 11 times since 2025, most recently on 22 October 2026.
Pros & Cons:
Pros: Fully local processing means audio and video never leave the machine, which matters for unreleased music, client footage, or confidential recordings; the free Basic tier includes batch processing and GPU acceleration rather than gating them behind Pro, so evaluation is not artificially limited; Pro at $19/month does not auto-renew, avoiding the subscription traps common in this category; it handles video dialogue and sound-effect separation alongside music stems, which many stem splitters do not; the vendor’s technical blog includes Demucs code examples and model comparisons (Becruily Mel-Roformer, BS-Roformer), suggesting real domain knowledge rather than a thin wrapper; RightAIChoice has repeatedly re-verified pricing and features.
Cons: **Independent user reviews are extremely thin**—RightAIChoice tracks only one source (Product Hunt) with a 60% positive / 40% critical split and explicitly notes the absence of community reviews or testimonials to validate the “perfect separation” claims; **company information is opaque**—WHOIS uses a paid privacy service, registration points to Jiangsu, and the operating entity, team, and funding are undisclosed, with ScamAdviser noting the site had not been scanned for over 30 days; **GPU support is limited to NVIDIA M250 and above**, excluding AMD and leaving Apple Silicon Mac users on slow CPU processing; **it is a local desktop tool** with no API, no cloud collaboration, and no browser version, so it does not serve users who want to open a web page and start immediately; **free-tier and Pro model quality differ**, with top-tier vocal and film models reserved for Pro.
**Who is it for?**:Producers and musicians who need to isolate vocals or instruments from existing tracks without uploading unreleased material; video editors separating dialogue, music, and effects from footage; podcasters and interviewers cleaning ambient noise from recordings; privacy-conscious users with NVIDIA GPUs who prefer local processing over cloud services.