Netwrck
A Built-in Image Editor and a Multitrack Audio Studio
Two things changed in this release. Editing an image no longer sends you to a separate site, and audio work is no longer limited to one file at a time. There is also a new multitrack studio with stem separation and AI generation built in.
Open Audio Studio Try the image editor in the gallery
The image editor is now part of the product
Every image generator previously had an Open in Editor button that opened a different website in a new tab. Your generation history, your account, and your credits stayed behind. That button now opens an editor built into the page.
It is a full-resolution canvas editor. Adjustments run as a single-pass WebGL shader, so a seventeen-slider stack repaints as fast as you can drag, and there is a 2D-canvas fallback for machines without WebGL.
What it does
- Adjust: exposure, contrast, highlights, shadows, whites, blacks, saturation, vibrance, temperature, tint, hue, clarity, sharpen, blur, fade, vignette and grain. Auto-enhance reads the image histogram and picks a starting point.
- Filters: fifteen presets, each rendered as a live thumbnail of your actual image rather than a generic swatch. Presets set the sliders, so you can keep tuning afterwards.
- Tools: crop with nine aspect presets, rotate, flip, paint, pixel erase, text with outlines, rectangles, ellipses, arrows, and an eyedropper.
- Masking: a selection brush that marks where AI edits land.
- History: undo and redo across every operation, plus hold-to-compare against the original.
Masked generative editing
The AI panel runs generative edit, background removal, upscaling and relighting. The masking behaviour is worth calling out: if you brush a region before running a generative edit, only the brushed pixels are replaced, composited back with a feathered edge. Without a mask, the model output replaces the whole image.
This means you can change one object in a scene without the model quietly redrawing the rest of it.
| Operation | Model | Credits |
|---|---|---|
| Generative edit | Qwen Image Edit | 8 |
| Fast edit | Seedream v4.5 Edit | 4 |
| Remove background | BiRefNet v2 | 2 |
| Upscale 2x | Creative Upscaler | 15 |
| Relight | IC-Light v2 | 8 |
Where you will find it
The editor is available on every tool page, on gallery tiles, and on post pages. Finished images download, copy to the clipboard, or post straight to the gallery. Inside the video editor, edited stills save back into your asset library.
Audio Studio: multitrack, with the library attached
Audio Studio is a new tool. It is a timeline with as many tracks as you need, clips you drag and trim, and per-track volume, pan, mute and solo. Playback runs through the Web Audio API so scrubbing is immediate; the final mixdown is rendered server-side with FFmpeg so the export is exact.
- Drag clips along the timeline, trim from either edge, split at the playhead
- BPM grid with snapping, zoom, loop
- Per-track gain and constant-power pan, mute and solo
- Mixdown to MP3, WAV, FLAC, M4A, OGG or Opus with a limiter on the master bus
Everything you need is in the sidebar
Rather than making you generate audio elsewhere and upload it, the panels sit next to the timeline:
- Files: drag and drop uploads
- Library: search the royalty-free music, SFX and ambience index, preview before importing
- AI Music: describe a track and generate it onto the timeline
- AI SFX: text-to-sound-effect
- Voice: text to speech, with optional reference audio from Voice Clone
- Stems: split an existing track apart
Stem separation, three ways
Separation runs on whichever engine is configured, and the interface tells you which are available rather than failing at submit time.
| Engine | Stems | Cost | Quality |
|---|---|---|---|
| Fast split (FFmpeg) | instrumental, vocals, bass, mid, high | Free | Approximate |
| Demucs (native) | vocals, drums, bass, other | Free | High |
| Hosted model | vocals, drums, bass, other | Credits | High |
The FFmpeg engine is honest about what it is. It uses mid/side cancellation and band splitting, which produces a genuinely useful instrumental on most stereo mixes and an approximate vocal track. It is instant and free. For true four-stem separation, use a neural engine.
Separated stems arrive as new tracks on the timeline, ready to remix.
The single-file audio editor got the same treatment
The original Audio Editor is still there for quick one-file cleanups, and it is faster to use than the studio for that job. It is no longer a wall of numeric inputs:
- An interactive waveform. Drag across it to set a trim range, click to seek. The selection is shaded and the playhead tracks playback.
- Nine one-click presets: Voice cleanup, Podcast, Music master, Loudness only, Denoise only, Phone call, Radio warmth, Bright and clear, and Reset.
- AI generation straight into the editor, so you can generate a track and immediately clean and export it.
- Stem separation inline, with each stem playable, downloadable, or promotable to the working file.
The underlying processing is unchanged: FFmpeg denoising, de-click, de-clip, EQ, compression, de-essing, silence trimming, loudness normalisation and limiting.
Video editor
The video editor keeps its timeline and stage, with the AI workflow tightened up:
- Pressing Enter in a prompt field generates, instead of doing nothing
- Prompt suggestions under the art and video fields
- Every generated still gets an Edit button that opens the image editor and saves the result back to your assets
- The stage Edit action opens the built-in editor rather than a third-party site
- A link through to Audio Studio for building a soundtrack
Notes for API users
Two endpoints are new and generally useful:
GET /api/img-proxy?url=re-serves an allowlisted remote image same-origin so canvases can read it without tainting.POST /api/image-editor/uploadtakes base64 image bytes and returns a hosted URL, which you can then pass to any endpoint that accepts animage_url.
Audio Studio is backed by /api/audio-studio/backends, /peaks, /ingest, /separate and /render. The render endpoint takes the full track and clip model as JSON and streams back the mixed file.