replying to @dtinth
From my preliminary usage, it tends to transcribe numbers, like spoken numbers, as words rather than as digits. It also tends to spell out English words in Thai when it is placed inside a Thai sentence.
Commit
461c55a
@dtinth Polishing = disfluency_removal_enabled. Best: language_hints ["th"] + disfluency_removal_enabled: Latin words kept 12/12 (default ~4 in 10 phonetic), numbers stay digits. A hint alone makes numbers words. Want me to ship that as our provider default? Thai numbers…
Please just enable
disfluency_removal_enabled by default. Make a PR pls.@dtinth Two draft PRs, checks green: #127 polishing on by default (message + streaming, opt-out config), #126 transcripts through @thai/text-cleaner everywhere. Streaming gains word spaces too. Not marked ready or merged; Thai numbers untested.
Combine into 1 PR and mark ready pls.
Combined into #127 (text-cleaner + polishing default); #126 closed. 486 tests pass. Ready is blocked: please /permissions allow 'claw exec -- gh pr ready 127 --repo dtinth/vxbeamer' @dtinth
Read both. Grok STT 2.0: $0.10/h batch, $0.20/h streaming (OpenRouter $0.000028/s=$0.10/h), word timestamps, key terms (100), filler removal, 19+ langs, auto language. No xAI provider in vxasr yet. Want me to add one (streaming) + test Thai? Need XAI_API_KEY @dtinth
Yes please. XAI_API_KEY added to .env
Draft PR #128: xai/grok-voice-transcribe-2.0, partials ~1s, $0.20/h. Formatting on (needs language, Thai unsupported, so en; Thai numbers untested). Server stalls 10-40s sometimes. @dtinth
Please make an infographic that shows the model behavior
Grok infographic: partials from ~1s, text self-corrects, final is the utterance event, end wait stalls 10-40s sometimes, formatting only helps with en. @dtinth
I want to try out Scribe Realtime v2.
https://elevenlabs.io/docs/eleven-api/guides/how-to/speech-to-text/realtime/server-side-streaming
Want to see similar infographic.
Also want to know whether it supports faster than realtime dumping.
elevenlabs.ioServer-side streaming | ElevenLabs DocumentationThis guide shows you how to transcribe audio in real time on the server side using ElevenLabs.
On it. No ElevenLabs key in .env yet: please add ELEVENLABS_API_KEY. Reading the docs for the model id, pricing and speed meanwhile. @dtinth
API key added
Scribe v2 Realtime: yes, faster-than-realtime dump works (9s in 1.6s, 37s in 2.7s, same text). 2.2s to first partial, 0.3s commit-to-final, $0.22/h. Add a provider (draft PR)? @dtinth
yes please
Draft PR #130: elevenlabs/scribe_v2_realtime, fast-dump capable, 500 tests pass. Handles 0.3s commit minimum + vendor auto-commit race. Conflicts with #128 on lists. @dtinth
Please combine 128 and 130 into a single PR, shepherd to green, then merge.
After that, I want to test this: https://iapp.co.th/docs/speech/speech-to-text/realtime
iapp.co.thOpenThai ASR Realtime | iApp TechnologyOpenThai ASR Realtime ถอดเสียงพูดภาษาไทยเป็นข้อความแบบสดผ่าน WebSocket: ได้ประโยคที่พูดมาถึงตอนนั้นราวทุก 0.7 วินาทีระหว่างพูด และได้ประโยคที่จบแล้วโดยค่ากลาง 0.66 วินาทีหลังผู้พูดหยุด ใช้ฟรีในช่วงเปิดตัว
commit 461c55a
Combined into #128 (130 closed), CI green, marked ready. Merge is blocked: please /permissions allow 'claw exec -- gh pr merge 128 --repo dtinth/vxbeamer --rebase'. Starting on iApp meanwhile. @dtinth
iApp realtime-v2: Thai-only, free at launch, apikey header. Their docs say English words come back in Thai script; fast dump ok up to 2 min ahead. Need a key: please add IAPP_API_KEY to .env. @dtinth
#128 merged (rebase) into main: xai Grok + ElevenLabs Scribe v2 providers, CI was green. Still need IAPP_API_KEY in .env to start the iApp realtime test. @dtinth