Repository navigation
Stop the Models card misreading as GPU memory, rewrite the README, add a social preview - #13
Merged
Merged
Conversation
The hardware card at the top of Models set the GPU's video memory in small grey
type ("8 GB") and, beside it, a bold "12.6 GB" that was the speech models' size
on disk. In the README's GIF that read as the card's memory: the owner's GTX 1070
has 8 GB, and the bold number looked like the headline.
The card is now about the GPU alone, with its memory labelled ("8 GB video
memory"). The disk figure and Open folder move to the Speech recognition card's
header, above the table that lists each model's size, as "12.6 GB on this PC".
Checked in headless Edge at 1100x620, 1200x800, 900 and 720 wide: the GPU row
fits on one line at all four, where the old right-hand column squeezed its badges
onto a second line.
docs/flow.gif is re-recorded with scripts/home_reel.py so the README shows it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The owner found the README accurate but stiff, like generated text rather than something written by hand. Checked against Wikipedia's "Signs of AI writing", Art of README, makeareadme.com and the READMEs of Handy, VoiceInk and Maccy, the old one had the tells that page lists: bullet lists that open on a bold phrase and a colon, em dashes where a comma would do, bold scattered through the prose, and a tagline that was three facts joined by dots. It also opened on a GIF caption about how the GIF was recorded. The new one: - opens like the apps above: the icon, the name, one sentence on what Flow is and who it's for, four badges (release, Windows 10 | 11, CI, licence), and links to install, the guide and the issue tracker; - says in plain paragraphs, with contractions, what happens when you hold the keys, what the pill's other two modes do, and who Flow was made for; - leads the install with the zip for Windows, then Scoop, then uv, with the SmartScreen step as GitHub's one Note alert, and gives macOS and Linux (Lite) their own subsection; - sets the three modes out as a table, keeps the privacy and known-limits lists as plain sentences, and turns the docs table into one paragraph; - moves the line about scripts/home_reel.py down to Contributing, and ends with the licence and the projects Flow is built on. Every command, link and figure the old one carried is still there, including the measured word-list numbers. tests/test_packaging.py found the tour by "## The loop"; it now finds "## Using it", and fails if that heading moves again instead of passing vacuously. docs/guide.md's link to the README's Lite section follows its new anchor. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Asked for by the owner: the header named only Windows, so the Lite build for macOS
and Linux was invisible until the install section. It gets a badge beside Windows'
("macOS | Linux: Lite, in progress", in the amber Flow uses for in-between states)
that links to that section, and the section itself now opens by saying Lite is
still in progress. It does not promise paste or global hotkeys off Windows:
product.md says a native macOS body is not promised.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…om the reel In the README's GIF the pointer stayed where it clicked the composer. Once the question posts, the composer drops below the answer, so the answer arrived under the arrow and stayed there for three seconds. After Enter the hand now glides to the empty space under the conversation list and rests there while the answer comes. The pause is the same length, so the reel's timing is unchanged; docs/flow.gif is re-recorded with it. scripts/social_card.py builds the repo's social preview from that reel: a 1280x640 card with the words on the left, in Flow Home's fonts and colours with the same Windows and "macOS · Linux: Lite, in progress" chips as the README's badges, and the reel at 80 % on the right, down to the pill in its desk strip. - **First frame:** X, Slack and LinkedIn show only a GIF's first frame, so the card opens on the answered question with the pointer at rest. Where GIFs move, it plays the tour on from there. - **Size:** GitHub takes a social preview under 1 MB. This one is 864 KB, and the script fails if a future reel pushes it over. - **Not committed:** GitHub has no API for the social preview, so it is uploaded by hand and GitHub keeps its own copy. docs/social-preview.gif is git-ignored, and that also keeps it out of the sdist, which it alone put over its 2 MB guard. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
samartomar
added a commit
that referenced
this pull request
Sep 24, 2026
…gline (#14) #13 gave the README header and the social preview a new tagline ("Local voice dictation for Windows, made for developers who speak English with an accent") while the owner chose to keep the About description, so the repo page showed two different one-liners. Both now say what the About box and pyproject.toml already say: "Local English dictation with a talk-to-it refine loop". Who Flow is for is still the README's third paragraph. On the card the line breaks after "dictation", so that "talk-to-it refine loop" stays together, and it is set a little larger, at 30 px, now that it is two lines rather than three. The block moves down to stay centred. Co-authored-by: samartomar <9993940+samartomar@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Four changes that make the repo's front page read better, one per commit.
The Models card no longer makes the disk size look like the GPU's memory. In the README's GIF, the GPU card showed "8 GB" small and grey, with a bold "12.6 GB" beside it that was actually the speech models' size on disk. The card is now about the GPU alone ("8 GB video memory"). The disk figure and Open folder moved to the Speech recognition card's header, above the table of model sizes, as "12.6 GB on this PC". I checked it in headless Edge at 1100, 1200, 900 and 720 px wide, and the GPU row now fits on one line at every width.
The README is rewritten to read like a person wrote it. I checked it against Wikipedia's Signs of AI writing, Art of README, makeareadme.com and the READMEs of Handy, VoiceInk and Maccy. The old one had most of the tells those list:
What the new one does:
test_packagingnow finds the tour at "## Using it", and fails if that heading moves, rather than passing without it.macOS and Linux are marked "Lite, in progress" with an amber badge next to Windows' badge, and the Lite section says so too. It doesn't promise paste or global hotkeys off Windows, because product.md doesn't.
The reel keeps its pointer off the answer, and a new script makes a social preview from the reel.
docs/flow.gifis re-recorded.scripts/social_card.pymakes a 1280×640 card, with the reel playing on the right.docs/social-preview.gifis git-ignored. That also keeps it out of the sdist: the card alone put the sdist over its 2 MB guard.The suite has 3205 tests. Locally, the only failures are the 13 known
test_refineones, from this machine'scodex.CMDshim.🤖 Generated with Claude Code