Can I run the whole setup on a laptop with no GPU?

No. The ASR side needs an NVIDIA GPU with ample memory behind vLLM plus model and VAD weights, while the companion server itself is stdlib-only Python and the page is a single static file.

Is Chinese annotation as reliable as English here?

No, English is stronger. Low-confidence Chinese results are left unmarked, and adding an English description per label helps noticeably; thresholds are tunable constants.

Do live labels flicker or fight with final labels?

Live labels are debounced and rate-limited, then the closed-sentence final judgment overwrites them; append-only ASR text never rewrites, so annotations can safely stick to sentences.

Where do my TypeSafe and DeepSeek keys actually go?

In browser localStorage, sent through a same-origin proxy so the server never writes them to disk. Public deploys still need HTTPS for microphone access.

What does the semantic-live-caption pipeline combine?

Confucius4-R2T2 for streaming ASR, TypeSafe Jev for per-sentence real-time judgments, and optionally DeepSeek to discover new labels; it targets meetings, interviews, podcasts and live streams.

What are the three annotation dimensions?

Emphasis (bold plus a wavy underline and highlighter), emotion (a superscript note in kai script plus a sentence background color), and intent (a stamp such as question, action, agreement or objection).

What gates must an intent label pass?

Four doors: a choice hit, confidence at or above 0.55, a lead margin of at least 0.12, and a Noul recheck, with high-false-positive categories like objection requiring 0.70.

CommunityJev Tools & CLIs

Semantic Live Caption

Streaming caption page pairing append-only ASR with per-sentence Jev annotation for key points, emotion, and intent, with custom labels and optional DeepSeek discovery.

See highlights

Open github.com

Stars: 10
Language: HTML
License: MIT
Published by: Community
Author: limboinf
Last verified: Oct 2026

limboinf/semantic-live-caption

User repository on GitHub

View on GitHub

听写纸 · Real-time speech captions with live semantic annotation (key points / emotion / intent) — Confucius4-R2T2 + TypeSafe Jev + DeepSeek

Forks: 1
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity awesomejev catalog-import html

Works forJev for Classification Jev for Routing

Used inJev for Software & DevTools

Highlights

Labels partial speech at live level, then overwrites with final judgment on close.
Gates intent with choice hit, confidence, margin, and Noul recheck plus strict lines.
Runs as one dependency-free HTML file with keys in localStorage via proxy.
Grows labels via DeepSeek suggestions with adopt or ignore persisted.

Watch out

ASR needs a GPU server with Linux and Python, and Chinese calibration is weaker than English so low-confidence labels are skipped. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

jevclip

Video tool where Jev scores subtitle segments and code assembles a short highlights cut plus a de-watered full cut with timestamped summaries.

Jev Tools & CLIs#community#jevbest#catalog-import

Community

cclank/jevclipPython

jevseek

Experiment turning Jev into a sampler: DeepSeek proposes top-k next tokens with logprobs and Jev picks each token, also judging when to stop or cut rambling.

Jev Tools & CLIs#community#awesomejev#catalog-import

Community

blingdivinity/jevseekPython

jev-resume-disqualifier

Recruiter dashboard that applies calendar date math plus visa, tenure, license and clearance gates, then a Jev decision step, sorting resumes into disqualified, advance or manual review.

Jev Tools & CLIs#community#jevbest#catalog-import

Community

AiPersonacademy/jev-resume-disqualifierPython

jev-trade-cc

A point-in-time backtest of a William O'Neil momentum strategy where jev picks the entries and decides the exits, paired against the same rules with no model — and honest enough to report the model as indistinguishable from them.

Jev Tools & CLIs#community#awesomejev#catalog-import

Community

michaelpersonal/jev-trade-ccPython

jackalope

A Tauri desktop workspace for running coding agents side by side: parallel task queues in isolated Git worktrees, your existing agent sign-ins, and a review step before anything is merged.

Jev Tools & CLIs#community#jevbest#catalog-import

Community

Jackalope-Dev/jackalopeRust

call-coach-ai

Desktop and web live-call coach that sends transcript lines to TypeSafe Jev after each sentence and shows a 0-100 gauge with a suggested next step.

Jev Tools & CLIs#community#jevbest#catalog-import

Community

ZeroGold/call-coach-aiHTML

From the Jev guides

Original write-ups on what this entry is for.

Jev on the Gateways: Vercel, Workers AI, and OpenRouter

Oct 12, 2026

The request shape barely changes across surfaces; the surrounding contract does. When to call the API directly and when a gateway, a Worker binding, or a router entry earns its extra hop.

Jev Tools & CLIs#gateway#vercel#cloudflare

Read article

Exporting Decision Models: GGUF, ONNX, and Device Runtimes

Oct 10, 2026

Taking a decision checkpoint out of Python: what the GGUF and browser WebGPU exports preserve, what they change, and the parity evidence the model cards publish.

Jev Tools & CLIs#open-models#local#onnx

Read article

Records at Rest: Documents, Inboxes, and Risk Queues

Sep 28, 2026

Jev labels a record that already exists. Code moves, flags, or holds it. How the document, mail, and transfer tools keep the action out of the model.

Jev Tools & CLIs#classification#email#security

Read article

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai