0632f7843b
Renumber spec 001 decision placeholders to D-040–D-051; record the Danish publication
Lucas WintherandClaude Opus 5.52026-10-30 01:47:27 +01:00
30969e86b7
Merge feat/parakeet-engine: Parakeet TDT 0.6B v3 English engine (spec 001 workstream 7)
Lucas Winther2026-10-30 01:46:51 +01:00
9c89ab047f
Merge feat/longform-pipeline: VAD long-form pipeline, opt-in (spec 001 workstream 3)
Lucas Winther2026-10-30 01:39:17 +01:00
19228297f1
Record the Danish conversions' measurements, dtype choice and pending upload in D-WS6a
Lucas WintherandClaude Opus 5.52026-10-30 01:13:27 +01:00
cc8e031d6b
Run WASM sessions at the default graph optimization level now that ort-web 1.31 loads the 8-bit decoders
Lucas WintherandClaude Opus 5.52026-10-30 00:45:46 +01:00
104c1c4e4d
Record the batched-VAD measurements and check heartbeat, bursts and cancel against the real sidecar
Lucas WintherandClaude Opus 5.52026-10-30 00:23:55 +01:00
1c1806840e
Tune the server engine's Silero VAD so batched file transcription stops dropping quiet speech
Lucas WintherandClaude Opus 5.52026-10-30 00:07:05 +01:00
378aff6a7b
Record D-WS2a: hint measurements, the 120-token truncation rule, no faster-whisper hotwords
Lucas WintherandClaude Opus 5.52026-10-29 23:52:29 +01:00
fdc9b500c0
Check Parakeet chunks export to .srt and timestamped text, and record the headless Chrome run
Lucas WintherandClaude Opus 5.52026-10-29 23:52:05 +01:00
9f1d58c0b7
Keep max_new_tokens: 440 on Transformers.js 4.3.2 and pin the seek-loop condition it avoids
Lucas WintherandClaude Opus 5.52026-10-29 22:59:35 +01:00
9cd92b4920
Record D-WS7a: Parakeet runtime decision, engine path and CPU measurements vs Large v3 Turbo
Lucas WintherandClaude Opus 5.52026-10-29 22:32:22 +01:00
cd9aa48b83
Record what the Transformers.js 4.3.2 upgrade changes on the WebGPU side and what needs a browser check
Lucas WintherandClaude Opus 5.52026-10-29 22:30:29 +01:00
12e0f6678b
Offer Parakeet in the picker as English-only, hide it on phones, and explain the missing language controls
Lucas WintherandClaude Opus 5.52026-10-29 22:08:08 +01:00
708fae5778
Run Parakeet TDT in the inference worker via parakeet.js on the shared onnxruntime-web
Lucas WintherandClaude Opus 5.52026-10-29 22:08:05 +01:00
f55ea6ecf8
Run browser file transcription through the long-form pipeline when asked and the VAD model is available
Lucas WintherandClaude Opus 5.52026-10-29 22:02:47 +01:00
2a4f5b16a5
Add Danish fine-tune catalog entries under a single pending Hugging Face owner constant
Lucas WintherandClaude Opus 5.52026-10-29 22:00:33 +01:00
f7d0525423
Wait for the PCM payload, not just the request, in the live WebSocket test (race under load)
Lucas WintherandClaude Opus 5.52026-10-29 21:58:03 +01:00
8236ea3504
End the hint prompt in a full stop: an unterminated list made Whisper drop punctuation and loop
Lucas WintherandClaude Opus 5.52026-10-29 21:55:48 +01:00
c5a57cd33a
Keep Parakeet on WASM unless WebGPU is chosen explicitly, and never offer it on phones
Lucas WintherandClaude Opus 5.52026-10-29 21:51:26 +01:00
827e9a8dae
Add a Cache-API model file downloader that reports load progress and caches only complete files
Lucas WintherandClaude Opus 5.52026-10-29 21:50:40 +01:00
c9fdab1726
Add Parakeet window planning, word-to-chunk mapping, vocab tokenizer and pinned asset URLs
Lucas WintherandClaude Opus 5.52026-10-29 21:49:52 +01:00
1c5952f212
Validate the vocabulary hint server-side and pass it to faster-whisper as initial_prompt
Lucas WintherandClaude Opus 5.52026-10-29 21:45:39 +01:00
313892b6df
Carry the vocabulary hint as prompt from the engines through the worker into every window
Lucas WintherandClaude Opus 5.52026-10-29 21:44:06 +01:00
e6de097cfd
Feed Whisper a decoder prompt through the ASR pipeline and strip it from the output
Lucas WintherandClaude Opus 5.52026-10-29 21:43:11 +01:00
afce794cbf
Mark server-live sidecar passes as live mode and cover the stream heartbeat with a test
Lucas WintherandClaude Opus 5.52026-10-29 21:37:01 +01:00
dc64bb8198
Run server-engine file transcriptions through faster-whisper's batched pipeline with VAD
Lucas WintherandClaude Opus 5.52026-10-29 21:36:21 +01:00
7ab41f4299
Document D-035: whisper-tiny multilingual vs .en speed gap is a content-dependent hallucination-loop effect, not a config bug
Lucas Winther2026-07-30 08:31:58 +02:00
500cf3b36e
Don't show the live "recording"/listening indicator until the model is ready
Lucas Winther2026-07-30 08:31:23 +02:00
c5d193600a
Show "Loading model from cache" instead of "Downloading" when weights are already cached
Lucas Winther2026-07-30 08:30:19 +02:00
bb77185e3f
Merge remote-tracking branch 'origin/main'
Lucas Winther2026-07-29 09:20:50 +02:00
7040f031a6
Usability and test infrastructure updates.
Lucas Winther2026-07-29 09:20:32 +02:00
