Changelog
Weekly highlights to keep you in sync with what changed.
Last updated: January 14, 2025
FeatureImproveStability
Apr 7 – Apr 13
Apr 7 – Apr 13
FeatureLatest
Gabox mainline switch completed with four new supplement model options
- 🔁 Mainline switch: Extract Vocals, Remove Vocals, and Karaoke now default to the new Gabox mainline models.
- ➕ New models: Added Gabox Fullness Voc, Gabox V8N Inst, Gabox Fullness V3 Inst, and Gabox Bleedless V3 Inst as selectable options.
- 🧾 History-safe display: Old Gabox mainline models were removed from the create-job list while historical jobs still keep friendly labels.
- 🌐 Multilingual sync: Model copy, changelog content, and the site banner were updated across supported languages.
Mar 17 – Mar 23
Mar 17 – Mar 23
Feature
Added two BS-RoFormer models with complete frontend support
- 🎤 New vocal model: Added BS Roformer Mag Anvuew to Extract Vocals for better magnitude-spectrum accuracy and vocal clarity.
- 🧼 New dereverb model: Added BS Roformer Dereverb Anvuew to Dereverb for cleaner noreverb output.
- 🌐 i18n update: Added localized model names and descriptions for consistent multilingual display.
- 🧩 Frontend sync: Updated model config and output stem typing so both models can be selected and submitted normally.
Jan 19 – Jan 25
Jan 19 – Jan 25
Feature
Payment Feature Launch & Service Model Adjustment
- 💳 Payment Launch: Official release of payment features (supporting PayPal / WeChat Pay).
- 📢 Service Model Adjustment: To balance GPU compute costs and ensure sustainable quality service, we have discontinued the daily free credit reset. In exchange, your credits are now permanently valid with no expiration. You can purchase and use them at your own pace.
- 🚀 Experience Improvements: Optimized pricing page visuals and enhanced core service stability and response speed.
Jan 12 – Jan 18
Jan 12 – Jan 18
Feature
Payment system & security upgrades, optimization of task workflow
- ✨ Payment: Redesigned pricing page with optimized workflow
- 🛡️ Security: Enforced account login security mechanism
- 🚀 Job Management: Improved queue stability and task synchronization
- 🎨 UI Polish: Updated console layout and download experience
- 🐛 Fixes: Fixed localization display and data processing issues
Jan 6 – Jan 12
Jan 6 – Jan 12
Feature
Payment system and account center launched, multi-model selection and Ensemble mode supported
- Payment system launched: Purchase audio processing packages online and track your orders and usage.
- Multi-model selection: Select multiple models to process the same audio simultaneously for more comprehensive results.
- Ensemble mode supported: Combine outputs from multiple models to overcome individual model limitations and significantly improve separation quality.
- Account center: View balance, order history, and usage statistics at a glance.
- Third-party login: More convenient account login options now available.
- Error page optimization: Brand new error pages with improved user experience.
- Added WeChat contact channel for Chinese users for faster customer support.
Dec 30 – Jan 5
Dec 30 – Jan 5
Improve
Model selection and download experience optimized
- Simplified model selection process to start tasks faster.
- Added one-click download link copy for easy result sharing.
- Optimized job processing logic for improved speed and stability.
Dec 23 – Dec 29
Dec 23 – Dec 29
Feature
Stability fixes, Hotspot Preview, new Scene pages, and FAQ center
- Fixed unresponsiveness for certain models during long runs, significantly improving system stability and reliability.
- Fixed issues causing excessive processing time for certain audio materials, improving overall processing efficiency.
- Fixed an issue where Live Cleanup did not generate the cleanmix stem.
- iPad download fix: resolved an issue where iPad could not download single-track files.
- Default output update: output-format now defaults to MP3 for broader compatibility.
- Smart model matching: removed quality tier selection; the system now matches the best-effect model by default.
- Model Update: Adopted the latest best-effect models, including 2stem (bs-roformer-anuvew), 4stem (scnet-xl-ihf), 6stem (bs-roformer-sw-6stem), and karaoke (melband-roformer-karaoke).
- Download filename optimization: File naming format changed to {original_name}_{track_type}.mp3 for better clarity.
- New Hotspot Preview: Automatically identifies audio highlights using smart algorithms to generate previews, helping you evaluate results faster.
- Launched standalone FAQ Help Center and Karaoke Maker scene pages for specialized tool access and better user support.
- Resources Explorer Preview: Introducing a dedicated browsing experience for our massive stem library (Work in Progress***).
- Job Retry: Retry failed jobs with one click, no need to re-upload files.
- Job Clone: Create new jobs based on existing ones, with optional scene/output adjustments.
- Detail Page Redesign: Redesigned job status display with intuitive retry/clone actions.
Dec 15 – Dec 21
Dec 15 – Dec 21
Feature
5.1 Surround Sound, New Cloud Storage & Stability Boost
- 5.1 Surround Sound Support: Added audio separation support for 5.1 surround sound, expanding professional audio use cases.
- Cloud Storage Upgrade: Migrated to a new cloud storage service for significantly faster and more stable file uploads.
- Stability Improvements: Fixed multiple known issues, improving overall system stability and responsiveness.
Dec 08 – Dec 14
Dec 08 – Dec 14
Improve
Experience Upgrade: Richer Track Types & Simpler Selection
- New Track Separation Types: Expanded support for diverse audio processing needs.
- Simpler Model Selection: Optimized job form with intuitive scene-based configuration to get you started faster.
- Optimized Progress Display: Improved job status and error messages for clearer tracking at a glance.
Dec 01 – Dec 07
Dec 01 – Dec 07
Feature
Major Update: "Scene Mode" & Scene Pages
- "Scene Mode" Launched: Our biggest usability upgrade. Skip complex parameters - just pick "Remove Vocals" or "Extract Vocals", and we'll match the best model.
- New Scene Pages: Launched dedicated landing pages (under Tools) for specific tasks like Vocal Removal and Extraction to help new users get started.
- Engine Upgrade: Optimized the job engine for better stability and speed at scale.
Nov 18 – Nov 24
Nov 18 – Nov 24
Feature
Dark Mode Refresh & Player Improvements
- Visual Upgrade: Unified dark mode design for a cleaner, focused interface.
- Easy Task Access: New "Recent Tasks" in the header to check history anytime.
- Better Playback: Streamlined audio player for smoother operation.
- Service Guarantee: Automatic refunds for failed tasks for worry-free usage.
Nov 11 – Nov 17
Nov 11 – Nov 17
Improve
Experience Optimization & Performance Boost
- New Subscription & Feedback: Stay updated and share your thoughts with us.
- Upgraded site architecture for faster loading and smoother interactions.
- Polished details like "back-to-top" behavior for a better browsing experience.
Nov 4 – Nov 10
Nov 4 – Nov 10
Feature
Multilingual Support & Player Upgrade
- Launched multilingual support (English/Chinese) with auto-detection.
- Upgraded the audio player for smoother preview and task management.
Oct 28 – Nov 3
Oct 28 – Nov 3
Stability
Keleeke initial launch
- Official launch: a cloud-based, zero-config professional vocal separation service built on open-source Keleeke models.
- Core Features: Support for vocal extraction, karaoke creation, and multiple audio format uploads.