draft 2026-10-21
last checked 2026-10-21
Making music without the screen: could an agent help?
21 October 2026. An open question, not a product claim.
Most music software is built for eyes. You see a track, click a clip, drag a fader, read a meter. Blind and low-vision producers make music anyway, and have for years, by working around tools that were not designed for them.
We build ismail, a music studio that an AI agent drives with text. It makes us wonder whether it could help here. We don't know yet, and this post is an invitation to find out.
How blind producers work today
A common setup is REAPER with OSARA, a free extension that makes REAPER usable with a screen reader on Windows and Mac. Pro Tools and Logic have built-in screen reader support, and Ableton Live supports screen readers too: its accessibility FAQ says it tests with VoiceOver, NVDA, Narrator and JAWS.
A 2020 interview study of 18 visually impaired audio workers (Saha and Piper, ASSETS 2020) found that the hard part is often speed, not access:
- Plug-ins whose controls are not labelled cannot be used at all.
- Waveform and EQ displays are not designed for screen readers.
- Comparing levels means hearing a list of numbers and holding them in memory. One participant called it a "huge cognitive load".
- Speech talking over the music while it plays gets in the way.
Why an agent might help
ismail was built for a different reader: an AI agent, which cannot hear the music. So ismail reads every mix back to the agent as text: levels per bar, drum patterns, chords, song structure, and comparisons against a reference record. The person listens and says what they want, and the agent writes the change.
That shape could suit someone who can't see the screen:
- You talk; the agent edits. No clicking through menus or finding a track among 70.
- Reports come after the music, not over it. The agent reads a render once it is done, so it doesn't speak over playback.
- The agent can compare for you. Instead of reading out a table, it could say something like "the bass is 4 dB louder in the chorus than in the verse" and ask what you want.
- It builds the screen you need, or none. ismail's agent builds interfaces on request, such as a phone page or a feedback page. It could build one that works with your screen reader, or skip screens entirely.
These are reasons to try, not results. No blind or low-vision musician has tested ismail yet, as far as we know.
What we don't know
- Whether the setup is accessible: installing ismail means using a coding agent such as Claude Code, Codex or Cursor in a terminal. Claude Code has an opt-in screen reader mode; we have not tested it with ismail.
- Whether text readouts help or just add more to listen to.
- Whether the sound is good enough for your music.
An invitation
If you make music with a screen reader and want to try it, we would like to hear what works and what doesn't, in your words. Open an issue on ismail's GitHub, or ask your agent:
Set me up with ismail: https://github.com/newsbubbles/ismail
ismail is free and open source. If it helps, we'll say so here, with your permission. If it doesn't, we'll say that too.
Sources, read 20 October 2026 and re-checked 22 October 2026: OSARA site; Saha and Piper, "Understanding Audio Production Practices of People with Vision Impairments", ASSETS 2020; Ableton accessibility FAQ; Claude Code accessibility docs.