andodeki/makepad
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
makepad/libs/ai/models/speech/src
History
Exact
Admin c83ce3617d ai: hub stems and word timing, GPU yield, quantised image models, fleet discovery 
Stems take any audio file and can answer Ogg stems; speech-to-text answers
word-timed transcription and batched forced alignment. A generation the
full model cannot fit next to the UI runs on the family's q4 variant; a
chat during a GPU yield is answered; a GPU failure fails the job instead of
a silent CPU fallback. Block-quantised image transformer on the GPU and a
24 GB-card image model entry. The fleet beacon goes to every interface's
broadcast and a hub on the gateway is found where no beacon reaches; the
LAN listener can be stopped. Tokenizer and G2P lexicon tables download with
their models instead of living in git.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-14 07:24:22 +02:00
..
bin ai: hub stems and word timing, GPU yield, quantised image models, fleet discovery 2026-10-14 07:24:22 +02:00
g2p ai: hub stems and word timing, GPU yield, quantised image models, fleet discovery 2026-10-14 07:24:22 +02:00
kokoro ai: the AI hub gives way to the UI, smaller models, vision and speech 2026-10-13 12:26:21 +02:00
whisper ai: the AI hub gives way to the UI, smaller models, vision and speech 2026-10-13 12:26:21 +02:00
convert.rs libs: ai hub/models/cuda, speech, chat_ui 2026-09-15 13:40:31 +02:00
gpt2_split.rs platform, apps: deterministic app capture, one app registry, makepad-app-* names, and the platform fixes the Stage editor found -- the remote service gains a virtual clock (--virtual-clock; /step advances whole frames and answers after they are presented, bounded to 20 s), /settled, widget id paths in /snap (q=path:), in-process MP4 capture on that clock (/cap/start|stop|pause|resume, no overlay in the file), --window WxH@scale, a synthetic cursor drawn into frames and --seed; with those flags off nothing changes. Passes that sample another pass's target stay dirty until their source has drawn (black first frames under parallel load), a shader still compiling marks its pass dirty, the Mac text shader gets its pass's camera on the first draw, and texture readbacks go to the requester that asked (try_take_texture_readbacks_for). Splash: obj.delete removes a field, for-loop keys compare equal to their @id, a loop ending a module no longer empties the stack, and a call-scope leak is closed. CEF: one exclusive profile per process, relaunch handled without a Chrome window, JS dialogs suppressed, popups become embedded tabs, one app bundle per build. Speech: Kokoro reports word timings, whisper aligns known text (align_text) and word times are no longer a token late. libs/app_registry is the one list of apps (the WM and Builder read it, Stage finds apps by id); every app crate in apps/ is package and binary makepad-app-<id>. AGENTS.md records the current agent workflow and the rule against rm on a possibly empty variable. 2026-09-29 01:55:13 +02:00
indextts.rs libs: ai hub/models/cuda, speech, chat_ui 2026-09-15 13:40:31 +02:00
indextts_bigvgan.rs libs/ai: one AI stack, replacing libs/ggml, llama, mlx, cuda, tts, voice2 and pbr_paint 2026-08-23 01:34:35 +02:00
indextts_bigvgan_cuda.rs libs/ai: one AI stack, replacing libs/ggml, llama, mlx, cuda, tts, voice2 and pbr_paint 2026-08-23 01:34:35 +02:00
indextts_campplus.rs libs/ai: one AI stack, replacing libs/ggml, llama, mlx, cuda, tts, voice2 and pbr_paint 2026-08-23 01:34:35 +02:00
indextts_codec.rs libs/ai: one AI stack, replacing libs/ggml, llama, mlx, cuda, tts, voice2 and pbr_paint 2026-08-23 01:34:35 +02:00
indextts_codec_cuda.rs libs/ai: one AI stack, replacing libs/ggml, llama, mlx, cuda, tts, voice2 and pbr_paint 2026-08-23 01:34:35 +02:00
indextts_gpt.rs libs: ai hub/models/cuda, speech, chat_ui 2026-09-15 13:40:31 +02:00
indextts_mel.rs libs/ai: one AI stack, replacing libs/ggml, llama, mlx, cuda, tts, voice2 and pbr_paint 2026-08-23 01:34:35 +02:00
indextts_pipeline.rs libs: ai hub/models/cuda, speech, chat_ui 2026-09-15 13:40:31 +02:00
indextts_s2mel.rs libs/ai: one AI stack, replacing libs/ggml, llama, mlx, cuda, tts, voice2 and pbr_paint 2026-08-23 01:34:35 +02:00
indextts_s2mel_cuda.rs libs/ai: one AI stack, replacing libs/ggml, llama, mlx, cuda, tts, voice2 and pbr_paint 2026-08-23 01:34:35 +02:00
indextts_tokenizer.rs platform, apps: deterministic app capture, one app registry, makepad-app-* names, and the platform fixes the Stage editor found -- the remote service gains a virtual clock (--virtual-clock; /step advances whole frames and answers after they are presented, bounded to 20 s), /settled, widget id paths in /snap (q=path:), in-process MP4 capture on that clock (/cap/start|stop|pause|resume, no overlay in the file), --window WxH@scale, a synthetic cursor drawn into frames and --seed; with those flags off nothing changes. Passes that sample another pass's target stay dirty until their source has drawn (black first frames under parallel load), a shader still compiling marks its pass dirty, the Mac text shader gets its pass's camera on the first draw, and texture readbacks go to the requester that asked (try_take_texture_readbacks_for). Splash: obj.delete removes a field, for-loop keys compare equal to their @id, a loop ending a module no longer empties the stack, and a call-scope leak is closed. CEF: one exclusive profile per process, relaunch handled without a Chrome window, JS dialogs suppressed, popups become embedded tabs, one app bundle per build. Speech: Kokoro reports word timings, whisper aligns known text (align_text) and word times are no longer a token late. libs/app_registry is the one list of apps (the WM and Builder read it, Stage finds apps by id); every app crate in apps/ is package and binary makepad-app-<id>. AGENTS.md records the current agent workflow and the rule against rm on a possibly empty variable. 2026-09-29 01:55:13 +02:00
indextts_w2v.rs libs/ai: one AI stack, replacing libs/ggml, llama, mlx, cuda, tts, voice2 and pbr_paint 2026-08-23 01:34:35 +02:00
kokoro.rs ai: hub stems and word timing, GPU yield, quantised image models, fleet discovery 2026-10-14 07:24:22 +02:00
lib.rs platform, widgets, video, draw, speech, cef: ScreenCap records without costing frames, deterministic steps wait for async loads, an unpaced batch-render beat, clipboard images and phoneme timings -- the Stage campaign's Makepad side. 2026-09-29 15:00:34 +02:00
tts.rs platform, widgets, video, draw, speech, cef: ScreenCap records without costing frames, deterministic steps wait for async loads, an unpaced batch-render beat, clipboard images and phoneme timings -- the Stage campaign's Makepad side. 2026-09-29 15:00:34 +02:00
vad.rs libs: ai hub/models/cuda, speech, chat_ui 2026-09-15 13:40:31 +02:00