Free browser DAWs in 2026: what Audiotool, BandLab, Soundtrap and Veena give you without paying

This is a desk review, not a test. We compared what each maker publishes about its own free plan, linked every page we used, and gave no scores. We'll add scores once we've used each one hands-on (see how we test and rank). The four are listed alphabetically.

How do the free plans compare?

"Not listed" and "None listed" mean the page we checked doesn't say. They don't mean the feature is missing.

Which one should you start with?

If you never want to hit a cap: Audiotool. It's the only one of the four that says its free plan has no track, length or feature caps. Audiotool makes that claim on its own comparison page, so treat it as the maker's claim until we've tested it.

If you work on your phone as much as your laptop: BandLab. Studio runs on the web and in the mobile app. You can export individual tracks as WAV or M4A, or as MIDI. The catch is 16 tracks on the free plan, and most of the AI tools (AutoMix, Audio-to-MIDI, Voice Cleaner and others) are Pro/Max only; Pro is $99 for the first year and $149 after that. Free does get SplitterAI, at 3 splits a day.

If you're making music with other people in real time: Soundtrap. Sharing by link, live editing and built-in text and video chat are its main pitch. Its features page also lists a piano roll for MIDI and unlimited cloud storage.

If you want AI to draft parts with you: Veena. Veena is built around an AI co-producer, and the free plan includes it. The CoProducer generates audio and MIDI inside the session, with a daily request allowance that resets at midnight UTC. The free plan also exports WAV (24-bit/48 kHz), MP3, MIDI and stems with no watermark, and runs VST3 and AU plug-ins through Veena Bridge. The limits are real, though: five projects, and only the basic AI models. Pro is $20 a month.

What we haven't checked yet

We haven't checked sound quality, latency, how stable each one is with a full session, or what each maker's terms say about who owns AI-generated output. Those need hands-on testing, and they'll decide the scores.

How we'll score them

When we test these four, we'll use the scoring rubric and its published weights, with the same rules for each:

• Performance (35). Sound quality, latency and how stable each one is with a full session.
• Rights and data (20). What each maker's terms say about who owns AI-generated output and what happens to your projects.
• Cost and limits (20). How far the free plan really goes, and what the first paid step costs.
• Workflow and formats (15). Plug-in support, collaboration, and export formats.
• Control (10). How much of a session you can shape by hand, including after any AI-generated part.

If a maker thinks we've got a published detail wrong, tell us and we'll correct it here.