Repository navigation
v0.4.0
What's Changed
- user-experience.md: phone-first exams lesson, one line reworded by @newsbubbles in #206
- Skill: the contributor dev, the dev role on the person's machine by @newsbubbles in #207
- Skill: rigged-characters.md, rigged bodies on the stage by @newsbubbles in #208
- video render takes the machine's GPU slot (hq:D-131) by @newsbubbles in #209
- Skill: loudness and windows references, live highlights, generic music video by @newsbubbles in #210
- Skill: lessons say "the person", not one user's name by @newsbubbles in #211
- Dev copies: python -m ismail.worktrees tracks every git worktree by @newsbubbles in #212
- Video: a text overlay, off by default (CLI + C.overlay) by @newsbubbles in #214
- Live ops: live_track output, live_start trim/cap and engine priority by @newsbubbles in #213
- Revert the video text overlay (pr:214) by @newsbubbles in #215
- phone_watch: one standing phone watcher with a cursor on disk by @newsbubbles in #216
- CHANGELOG: drop the leftover overlay entry by @newsbubbles in #217
- Skill: one piece, one folder, and moving a live engine to a new set's folder by @newsbubbles in #218
- Stage: nine small bugs from the engine review by @newsbubbles in #201
- Stage: people's mouths move with a voice (lip sync) by @newsbubbles in #205
- Phone tests: a stale test's transcriber no longer reaches the next test by @newsbubbles in #219
- Phone: still background when no set plays; phone_now past 200 marks by @newsbubbles in #220
- Skill: a piece copied from another song is a whole song by @newsbubbles in #221
- Skill: three paces (studio, set, jam); a jam is threaded by @newsbubbles in #222
- Skill: a jam runner recipe by @newsbubbles in #223
- Stage: a pick made while another menu closes opens its menu by @newsbubbles in #225
- Stage: "heard by" receipts, busy lines, a stuck warning, and an ear that keeps listening by @newsbubbles in #226
- Stage: break the link -> api import cycle (ear could not start) by @newsbubbles in #227
- README: a please-credit section (licence stays MIT) by @newsbubbles in #228
- Stage lip sync hears the voice with Rhubarb Lip Sync by @newsbubbles in #229
- Stage: step into a person (stage_embody, Step in) by @newsbubbles in #230
- Stage: grab by touch, phase 1 by @newsbubbles in #231
- Stage: worn parts come off (stage_actor_part) by @newsbubbles in #232
- Render: faders in the readout, an empty artist tag, wait in minutes by @newsbubbles in #233
- README: the short front page Nate approved by @newsbubbles in #235
- Listen together: the set, the person's voice and their notes on one clock by @newsbubbles in #236
- Live: played keys, a MIDI keyboard played through the set by @newsbubbles in #237
- test_machine: never leak the paused 300 MB child by @newsbubbles in #238
- live_played: keys on the beat the player heard; fit=True snaps a take to the 16ths by @newsbubbles in #239
- listen: words as they are said (listen_heard), sessions off the system drive, a free-space floor by @newsbubbles in #240
- Stage: edits apply in place, no reload by @newsbubbles in #242
- Live: on-screen keys that play into the running engine (live_keys_page) by @newsbubbles in #241
- Stage: the gesture map (bindings as data, three locked rows) by @newsbubbles in #244
- Stage gestures: check saved rows on read, keep the answers on by @newsbubbles in #245
- Stage: named ops for every live edit (asset install, world update, restart) and an optional guard hook by @newsbubbles in #246
- Guard hook fails open (missing file, guard errors) by @newsbubbles in #247
- stage-dev: restarts check presence, not the person's word by @newsbubbles in #248
- worktrees: sweep the empty folders removed copies leave; done says HELD by @newsbubbles in #249
- Ops refuse: personal data on screen, asks a standing grant covers, mic without the person's yes by @newsbubbles in #250
- Grab by touch: hand-sized things, let-go after an open-handed take, steadier count by @newsbubbles in #251
- Camera moves in VR: trails on the clock, and a move recorded like a take by @newsbubbles in #252
- Voices: toms, real acoustic toms (Big Rusty, CC0) hit anywhere, any way by @newsbubbles in #253
- Stage: a command sent right after a scene switch runs by @newsbubbles in #255
- Add stage_preflight: one check before a VR session or demo by @newsbubbles in #256
- Skill: voice pages open in the person's own browser and keep what they said by @newsbubbles in #257
- feedback_page: a page the person answers by voice or typing, kept on this computer by @newsbubbles in #258
- spatial: move a sound in 3D for speakers or headphones, studio and live (+ rig.py cache-hash fix) by @newsbubbles in #259
- spatial: ctc on speakers no longer moves a side sound to the other speaker by @newsbubbles in #260
- stage: replay a take as input, agent-asked test takes, keep the last 2 minutes in VR by @newsbubbles in #261
- sound_zoom, sound_zoom_diff and region_measure (ledger:M193) by @newsbubbles in #262
- phone_presence and phone_claim: who is listening, and who passes a note on by @newsbubbles in #263
- stage: cameras that follow, baked into keys for the render by @newsbubbles in #264
- master_cue: master a finished cue to a loudness target, measured (M207) by @newsbubbles in #265
- Standing grants: owners, routes, a still-ask list and grant_check (M209, M210) by @newsbubbles in #266
- phone: watcher heartbeat, a watcher that stops while the person is live is named once (M211) by @newsbubbles in #267
- exams: measure first, then spend the person's ear (M212) by @newsbubbles in #269
- setup: the VR step (allow lines from stage_preflight, pasted by the person) by @newsbubbles in #270
- composition: a melody that travels through instruments (S50) by @newsbubbles in #271
- Governor: MISDECLARED, HUNG, ORPHAN, the pagefile's drive, machine disk beyond songs/ by @newsbubbles in #268
- live: headroom, cutting a section over, the room speaker off and on by @newsbubbles in #272
- Fresh clone: the help line, ismail's MCP version, the plain-words steps (M215, M216, M181 F3) by @newsbubbles in #274
- stage: capture setup at a song time, and a keyed screen for the comp (M217) by @newsbubbles in #273
- sketch: feeling words in the first brief, the tune level (M181 F2, M213) by @newsbubbles in #276
- Standing grants: sourced first, routes stop at still-ask and senses, down_since on watcher_down by @newsbubbles in #277
- Phone: under each note, who took it (ledger:M219) by @newsbubbles in #278
- docs: a quick start the melody reader reads right (M214) by @newsbubbles in #279
- docs: Live DJ lessons in the live and phone references (ledger:M223) by @newsbubbles in #280
- phone: one beatbox classifier, relative brightness (ledger:M222) by @newsbubbles in #281
- analyze: pitch in cents per note and a windowed pitch contour (ledger:M225) by @newsbubbles in #282
- analyze_melody: a quick octave jump reads in its own octave by @newsbubbles in #284
- song_compress: a finished song's WAVs as FLAC, bit for bit (ledger:M221) by @newsbubbles in #285
- analyze_melody: a held note reads as one note (ledger:M226) by @newsbubbles in #286
- Governor: the macOS meter sees a job's memory (M166) by @newsbubbles in #287
- Sampler: several sounds on one pitch; phone_now label expires (ledger:M218) by @newsbubbles in #288
- phone: a panel left unsent keeps what was typed, said or picked (ledger:M227) by @newsbubbles in #293
- skill: every song can be made again from its folder (ledger:M228) by @newsbubbles in #294
- One reclaim folder at the root of the drive (ledger:M229) by @newsbubbles in #295
- Tests: two stage tests no longer race the Windows runner by @newsbubbles in #290
- Stage: screens in the room (stage_surface) by @newsbubbles in #291
- Stage: a camera crew's tools (stage_on_air, stage_pointer) by @newsbubbles in #292
- worktrees done: take a copy whose pull request was squash-merged by @newsbubbles in #289
- analyze_tuning: every part of a render against the tuning grid and key (ledger:M231) by @newsbubbles in #296
- analyze_tuning: read fast notes and their attacks (ledger:M231) by @newsbubbles in #297
- phone: a linked .md, .csv or .log opens as text by @newsbubbles in #298
- Skill: fetching a reference from a link by @newsbubbles in #299
- Skill: one reference for helper agents by @newsbubbles in #300
- Skill: "ytdl" means yt-dlp, and it is a Python library too by @newsbubbles in #301
- Replies read the live state: reference line, feedback Send button, hear-it block by @newsbubbles in #303
- Skill: lessons from a first-session test, and a coherence pass by @newsbubbles in #302
- Migrate from a song: transitions voice, speech on a rhythm, analyze_voice_heard by @newsbubbles in #304
- Synthesis unless the user asks for samples by @newsbubbles in #305
- Balance against the reference in every render with stems by @newsbubbles in #306
- The effects pass is part of the first draft by @newsbubbles in #307
- Shared pages end with a made-with-ismail footer (M254) by @newsbubbles in #308
- Remote listeners: Copy my answers, a shareable page with voice notes, feedback_paste (M255) by @newsbubbles in #309
- Part names checked with the person by ear (M253) by @newsbubbles in #310
- ear_pass, and notes that name a quality come back as measure steps (M252) by @newsbubbles in #311
- live.md: helpers compose during a set; the mixer counts automation and finished decks by @newsbubbles in #312
- Live decks: a new song starts clean; removing a deck mid-block no longer stops the set by @newsbubbles in #313
- backup.set_aside moves proj/ whole or not at all by @newsbubbles in #314
- render reads the true peak; noise_riser tops out at 10.8 kHz by @newsbubbles in #315
- Skill: metal recreation lessons (recreate.md) and vocal arrangement lessons (composition.md) by @newsbubbles in #316
- feedback_page(moments=True): mark moments while the song plays by @newsbubbles in #317
- Live: a cut takes the old deck off air on the cut bar by @newsbubbles in #318
- Skill: one listener's taste reads as examples, not rules by @newsbubbles in #319
- README: a picture tour of what ismail looks like by @newsbubbles in #320
- Live streaming: one set to several sites, and every site's chat by @newsbubbles in #321
- Stream: a visual page drawn from the set (D-196 slice B) by @newsbubbles in #322
- Stream: YouTube chat in realtime over streamList (D-196 item 2) by @newsbubbles in #323
- Stream: a new MC sees the room's opening; practice page hides refused lines by @newsbubbles in #324
- phone: older downloads behind a button, paged from a kept history by @newsbubbles in #325
- MIDI files in and out: midi_write and midi_read by @newsbubbles in #326
- README: plugins are not a shortfall by @newsbubbles in #328
- Stream the stage: an on-air page and GPU window capture by @newsbubbles in #327
- MCP role filter: serve only some roles' or shelves' ops by @newsbubbles in #329
- A shared request queue for streams: request_* ops (ledger:M290) by @newsbubbles in #330
- Chat commands by rule, votes and credit for requests (ledger:M291) by @newsbubbles in #332
- Picture rounds from every song on one server: a round registry (ledger:M300) by @newsbubbles in #331
- Frozen tracks, and placed audio on decks: track_freeze (ledger:M292) by @newsbubbles in #334
- What a deck costs, before it plays: deck_cost (ledger:M293) by @newsbubbles in #333
- MIDI out to synths and other apps, and MIDI clock (ledger:M113 step 3) by @newsbubbles in #335
- Live: change the tempo of a running set at a bar (live_tempo) by @newsbubbles in #243
- The kit page: every voice and effect by family, from data (ledger:M295) by @newsbubbles in #337
- machine: the board lock serves callers in arrival order (ledger:M303) by @newsbubbles in #338
- machine: waiting jobs share one CPU and GPU reading and wake when a job ends (ledger:M307) by @newsbubbles in #344
- Tell ledger: tell_* ops per song (ledger:M305) by @newsbubbles in #340
- analyze_structure: an empty stem group reads silent, not loud (ledger:M296) by @newsbubbles in #336
- mimic: cap the level part of per-note variation; name outlier takes (ledger:M301) by @newsbubbles in #339
- mimic: dense chords start fast; the attack map runs on a numpy STFT (ledger:M302) by @newsbubbles in #341
- Blind tests: real or fake per clip, climb back to one note, score the window (M299) by @newsbubbles in #342
- voice_help: level per velocity; analyze_sections names the window (M288) by @newsbubbles in #343
- Velocity sets the level of rhodes and emily (ledger:M310) by @newsbubbles in #345
- mimic plays phrases: perform=true for legato winds and strings (ledger:M308) by @newsbubbles in #346
- notes_clash: semitone rubs between parts and out-of-key notes (M282) by @newsbubbles in #348
- Live: fn='perform' voices play as performers, as in the studio (M284) by @newsbubbles in #349
- Reading a drifting record: thumb vs finger band periods, analyze_swing(follow=True) (ledger:M287 a, b) by @newsbubbles in #350
- mimic perform: vibrato by default where a profile has it, and phrase ends that fall by taper_db by @newsbubbles in #355
- live_parity: one-sample offsets read as the same take; per-track render speed (ledger:M285(a)(e)) by @newsbubbles in #351
- analyze_grid: the harmony votes on bar 1 two beats apart (M312) by @newsbubbles in #352
- melody_consensus: a lead line over a backing in the same stem (M314) by @newsbubbles in #353
- instrument_fit takes wait; a band 20 dB off costs extra and warns (ledger:M313, M315) by @newsbubbles in #354
- analysis: analyze_sections reads a relative path from the cwd first; with stems it names the loudest part per section (ledger:M281 a, b) by @newsbubbles in #356
- master_cue: true-peak catch as the last stage; mp3=True gets its own 1.2 dB catch (ledger:M280) by @newsbubbles in #357
- mimic: short notes no longer drift tens of cents off (ledger:M306) by @newsbubbles in #347
- mimic perform: fingered slurs by default, and a vib_delay lane (ledger:M322, M323) by @newsbubbles in #358
- voices: flute, a measured air-jet performer (ledger:M321) by @newsbubbles in #359
- phone: a long path in the DJ log wraps instead of widening the page by @newsbubbles in #360
- phone: a long path in a panel's text wraps too by @newsbubbles in #361
- mimic perform: each slur's move inside the margin Nate named (ledger:M327) by @newsbubbles in #362
- packaging: one release step for harness plugin stores (ChatGPT/Codex, Claude Code) (hq:D-213, D-214) by @newsbubbles in #363
- voices: trumpet performer from kit_winds (M331) by @newsbubbles in #364
- Packaging: an install works like a clone, ready for PyPI (M351) by @newsbubbles in #365
- Release 0.4.0 (stacked on #365) by @newsbubbles in #366
Full Changelog: v0.3.0...v0.4.0