Skip to content

t27f token_floor: letters in Outfit ExtraBold, captions in JetBrains Mono (Closes #10376) - #10378

Merged
gHashTag merged 1 commit into
masterfrom
claude/token-floor-fonts
Oct 18, 2026
Merged

gHashTag merged 1 commit into
masterfrom
claude/token-floor-fonts

Conversation

@gHashTag

Copy link
Copy Markdown
Owner

Closes #10376

What

The token_floor reel (#10357) now uses t27.ai's own fonts: Outfit ExtraBold for the particle letters, JetBrains Mono ExtraBold for the captions.

  • specs/t27f/outfit_atlas.t27 (new). A 800x16 coverage atlas of Outfit ExtraBold for the 39 glyphs the reel uses, with advance widths and left bearings from the font's hmtx. The header records the woff2, static-instance, raster and row-data hashes plus the commands that made them. Tests check that every glyph is one run of ink, that the cap height is 12 rows, that the ink widths match the glyf bounding boxes within 1.5 dots, and that the H, comma, period and hyphen sit on the right rows.
  • token_floor.t27. The 5x7 bitmap font is gone. Each label's dots now sit on a 12-row grid inside the same 7-pitch cap height, each glyph is as wide as its advance, and each dot's light is the outline's coverage at that point, so stroke edges are anti-aliased. The reel now has 20889 dots (about 2000 before), and the counter needs 125 particles. Buffers: MAX_DOTS 22528, MAX_SPLATS 45312, POOL 160, with a test that they fit. New tests pin glyph dot counts measured on the lab (W 120, I 36, period 9), keep every label inside the frame, and keep every line above the floor clear of the tube.
  • token_floor_mux.t27. Captions use JetBrains Mono ExtraBold, static wght=800 cuts of t27.ai's woff2 files. The Cyrillic subset has no digits, so the three number words use the Latin cut. The size rule uses the mono advance (600 per mille), and the fit test now includes the border. Every word fits at the full 124 px.
  • Seals for the three specs.

Measured on the Railway lab

  • Test reports: outfit_atlas 5/5, token_floor 20/20, token_floor_mux 10/10, reel_particles_mux 6/6, none vacuous. All three specs seal.
  • Render: 52 s single-threaded for 450 frames. A ReleaseFast build is bit-identical (frames sha256 7e0c6662...) and no faster.
  • Reel: 15.000 s, 1080x1920 at 30 fps, one x264 encode at crf 14 with aq-mode 3, 12.9 MB, -14.0 LUFS integrated, LRA 2.6. The caption band is yellow at 1.0, 3.0, 5.0, 7.5, 9.5 and 12.3 s and empty at 13.8 s.

The font files, the voice audio and the narration words are data. They stay on the lab and are not in this PR.

🤖 Generated with Claude Code

…Mono (Closes #10376)

The particle labels now sample a coverage atlas of t27.ai's Outfit ExtraBold
(specs/t27f/outfit_atlas.t27): 12 rows to the cap, glyphs as wide as their
advance, each dot lit by its coverage. 20889 dots against about 2000 before;
buffers resized and tested. Captions use t27.ai's JetBrains Mono ExtraBold,
the Latin subset for the digit words. Seals added for the three specs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@gHashTag
gHashTag enabled auto-merge (squash) October 18, 2026 10:16
@github-actions

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-10-18 10:18:14 UTC

Summary

Status Count
Total Open PRs 50
PRs with Failing Checks 49
PRs with All Checks Green 1
READY 0
FAILING 49
PENDING 0
NO CHECKS YET 0

These columns do not partition: 0 + 49 + 0 + 0 = 49, and there are 50 open PRs. A PR is being counted twice or not at all.

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=fad12492937a != manifest seal=87e5cbd3ad94.
    The committed NMSE numbers were certified against an older compiler.rs.
    Run scripts/reseal-check.sh locally for the two-step reseal command (advisory; not a merge gate).

@gHashTag
gHashTag merged commit 9969315 into master Oct 18, 2026
26 of 32 checks passed
gHashTag added a commit that referenced this pull request Oct 18, 2026
…or.t27 is left to its owner (Refs #10383)

The shard partition and the TF_SHARD/TF_SHARDS reader become a generic
spec any renderer opts into with three lines in its main, so this PR no
longer edits token_floor.t27, which is being rewritten in PR #10378.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

t27f token_floor reel: letters in Outfit ExtraBold, captions in JetBrains Mono (t27.ai's own fonts)

1 participant