yoheinakajima /
glance-speedlab
A local lab for timing and speeding up live-camera vision-language inference
- 52 stars
- 1 stars since 3 Oct
- Python
- Updated 28 days ago
- Advanced
- Image, video & audio
Running a vision-language model on a live camera feed is slow, and it is hard to know which tweak actually helps. Glance Speedlab is a research harness for that. It grabs the newest camera frame only when the model is ready, sends one request with several questions, and reports capture, transport, image-prefix, scoring and full-loop latency separately.
It sits on top of Glance, a local layer that turns open vision-language models into typed yes/no decisions. Speedlab keeps a PyTorch MPS reference path next to an optional MLX 8-bit scorer, so you can switch between them in the same UI and compare.
Its main value is the written record. Each change starts as a preregistered experiment with a hypothesis and a quality guardrail, and the README tables what was kept and rejected. Results cover one Apple M5 laptop, Qwen3-VL-2B and 4B, and a small fixed test suite. It is a research prototype, not a production monitor.
What it does
Per-stage latency reporting
Capture, transport, image prefix, scoring and full-loop time are measured separately, so you can see what dominates.
Reference and MLX backends
Compare Glance on PyTorch MPS with an opt-in MLX 8-bit scorer. The README reports it 25 to 28% faster on fresh frames, matching all 84 fixed-suite decisions.
Logged experiments, failures included
The README tables about 16 interventions with outcomes. Native multi-question requests were 2.4x faster; 4-bit scoring was rejected for probability drift.
Benchmark command
pnpm bench times a still image with warmup and iteration counts and writes JSON to stdout.
Local and private by default
Everything binds to loopback, frames stay in memory, and metrics are recorded only if you turn them on, as timing metadata without frames.
What you could build
- Measure where latency goes in a local webcam vision-language pipeline on a Mac.
- Compare PyTorch MPS and MLX 8-bit scoring for Qwen3-VL-2B on your own hardware.
- Reproduce or challenge the published speed and quality results with the preregistered experiment workflow.
- Test whether a smaller model tier or lower image-token cap keeps your decisions stable.
- Use the experiment log as a template for documenting performance work honestly.
Get started
Clone and set up Glance first, cache its model, then install Speedlab's Node dependencies and launch it. Open http://127.0.0.1:8787 in a camera-enabled browser; use pnpm launch:mlx after creating the MLX environment for the 8-bit path.
git clone https://github.com/yoheinakajima/glance.git
cd glance
uv sync
uv run glance doctor --json
uv run glance ask samples/dog.jpg "Is there a dog?"corepack enable
pnpm install --frozen-lockfile
cp .env.example .env
pnpm launchCommands are copied from the project's README. Read it on GitHub for the full setup.
More in Image, video & audio
ComfyUI
Node-graph app for generating images, video, audio and 3D with local AI models
MoneyPrinterTurbo
Turns a topic or keyword into a finished short video with script, voice and subtitles
whisper
Speech recognition model that transcribes and translates audio in many languages
VoiceStudio
Desktop app for cloning voices, dubbing video and dictating, run on your own machine
OpenMontage
Video production pipelines your AI coding assistant runs from a plain-language prompt
hyperframes
Turn HTML, CSS and animations into repeatable MP4 videos, built for coding agents
voicebox
Local voice studio for cloning voices, generating speech and dictating into any app
upscayl
Desktop app that enlarges low-resolution images using AI, on Linux, macOS and Windows
VoxCPM
Open text-to-speech model that makes, designs and clones voices in 30 languages
awesome-gpt-image-2
A library of 530+ GPT Image 2 prompts with templates and an agent skill