Aloud

What is Aloud?

Aloud is an announced macOS tool for turning spoken product feedback into a task plan a coding agent can execute. You hit record and talk through your app while the screen is visible; Aloud captures voice, screen, and a live transcript together, rewrites the messy speech into what you actually meant, drops the calls you reversed mid-sentence, and suggests screenshots wherever you said "this", "here", or "that button". The result is exported as one self-contained file you drop into Claude Code, Cursor, or Codex. Transcription uses Whisper on-device. As of 2026-08-20 the product is not yet available: the site's download button reads "First build is coming", and no pricing is published.

Key features

• Records voice, screen, and a live transcript in a single session, with the recorder itself kept out of the captured video
• On-device Whisper transcription — the site labels the live transcript "whisper · on-device"
• Rewrites the raw transcript into what you meant, keeping only the calls you stood by when you changed your mind mid-session
• Flags ambiguous deictic references ("this", "here", "that button") and suggests the matching screenshot for each
• Frame-accurate screenshot insertion from the recording, with cropping and −1s / −frame / +frame / +1s scrubbing to pick the exact frame
• Groups the session into a structured plan — the site's example shows 3 groups and 11 tasks — with a model tier (fast or reasoning) assignable per group
• Exports one self-contained file to hand to Claude Code, Cursor, or Codex
• Supports long sessions; the site states you can "talk for minutes or hours"

Who it's for

• Reviewing your own build out loud and handing the agent a precise, screenshot-backed prompt instead of a paragraph of vague notes
• Capturing a design or UI critique where most of the meaning lives in what you are pointing at rather than in the words
• Recording a long walkthrough of an app and getting back a grouped task plan rather than a raw transcript to re-read
• Keeping product feedback local when the recording contains customer data or unreleased UI, since transcription runs on-device

When not to use it

Not usable yet: as of 2026-08-20 Aloud has not shipped. The download call-to-action on aloud.sh reads "First build is coming", so this is an announced product to watch, not one to adopt. It is also macOS-only by design, and it is aimed at people who narrate their work — if you already write tight written specs, the transcription-and-disambiguation step is overhead rather than a gain.

FAQ