Skip to content
saturalabs
← Back to the journal
About the journal

Useful notes.
For curious ears.

A place to learn what to try, what to listen for, and when to leave a recording alone.

Published by Saturalabs

Saturalabs is a browser-based tool for isolating sounds from audio and video. This journal explains workflows for dialogue cleanup, music separation, and working with individual sounds. It is published by the product, so guides may link to Saturalabs tools where they fit the task.

Practical steps, with the limits included

Each launch guide focuses on a specific job and gives you a way to evaluate the result. Example prompts are starting points, not guarantees. A separated track is an estimate; the quality of the source and the overlap between sounds affect what you can recover.

How the launch guides were made

The initial guides were prepared with AI assistance using the current Saturalabs interface and workflow as references. Instructional steps are separate from the original signal comparisons described below. External documentation is linked where it supports a technical explanation; neither a citation nor a synthetic test is a claim of independent expert review.

Original examples, with methods you can inspect

Some guides include original synthetic signals and measured exports. We describe the source, processing settings and measurement method alongside those examples, and provide the actual files rather than an illustration of an imagined result. These comparisons test a defined operation; they do not establish how well AI will separate a real voice, instrument or noisy recording.

  • Channel and mono comparisons use known matching, one-sided and opposing-polarity signals to show what channel arithmetic keeps or cancels.
  • Equal-peak loudness examples distinguish the highest stored sample from integrated loudness, with measurement notes and independent FFmpeg checks.
  • Hum and filter examples compare known tone components before and after filtering. Those readings are not percentages of real-world noise removal.
  • Actual WAV and MP3 downloads let you inspect stored bytes, payload estimates and decoded duration, without treating a size ratio as a quality score.
  • A constructed chime, background and mixture let you compare known source components. Both components were synthesized before mixing; neither is an AI recovery from the combined file.
  • M4A and MP3 size references compare actual outputs from the app’s encoder in a native reference run, with stored byte sizes and decoded durations. They do not establish browser compatibility or predict the size or listening quality of another recording.

The linked generation notes and measurement reports identify the files and explain the checks. Where provided, checksums let you confirm which bytes a report describes. The WAV reference library collects additional original signals; reuse terms appear with each group. Native file checks do not certify every browser or playback device, and the synthetic examples are not substitutes for evaluating your own recording.

Product workflows and outside references

Instructions distinguish Saturalabs' account-and-credit AI service from its separate free local utilities. A local trim, channel extraction or frequency filter does not start an AI separation. When a guide discusses another product, its linked official documentation supports the stated feature or workflow; that alone does not mean we performed a hands-on comparison. Prices, free allowances and compatibility can change, so check the linked provider before choosing it for a project.

Dates and corrections

Publication dates show when a guide was added. A changed date should reflect a substantive update, such as revised instructions or an important correction. If something does not match your experience, contact saturalabs@protonmail.com with the guide title and the step that needs attention.

Start with your next project

Explore the guides or find the right starting point in our audio tools. Keep your original recording and judge a processed version in the context where you will use it.