Skip to content

Image Studio v0.2.9 (8GB VRAM Profiles, Immediate UI Launch & Large Image Fix)

Choose a tag to compare

View all tags
@DeSinc DeSinc released this 12 Oct 07:43
· 32 commits to main since this release
v0.2.9
9dc4102

Image Studio v0.2.9

What's New in v0.2.9:

  • Large Image Upload Fix:
    • Fixed Error: upload reply: error decoding response body when adding high-resolution images (e.g. 14,000px+, 25MB+).
    • Reference images are intelligently scaled to a maximum of 2048px (preserving aspect ratio), preventing aiohttp 100MB body limit rejections (HTTP 413) and vision transformer CUDA OOM.
    • Base images are scaled to max 4096px (diffusion ceiling and GPU texture limit).
    • Explicit HTTP response error reporting in comfy::upload.
  • 8GB VRAM Quantized Models & Hardware Auto-Detection:
    • Added Qwen Image 2.1 (8GB VRAM - W4A8 Quant) profile utilizing 4-bit quantized text encoder (qwen3vl_8b_w4a8.safetensors, 6.31GB) to prevent CUDA Out Of Memory crashes on 8GB GPUs.
    • Added Turbo 8-Step (8GB VRAM) profile for accelerated inference.
    • Automatic GPU VRAM detection: systems with <= 8GB VRAM automatically select the 8GB profile, while systems with 12GB+ use the standard profile.
    • Profile switcher in the Setup & diagnostics window allows manual selection with automatic weight downloading.
  • Immediate Main UI Launch & Non-Blocking Startup:
    • The full main UI and hero canvas are now shown immediately upon launch when starting the engine (no blocking setup card).
    • Top-right header shows non-intrusive status (Starting engine...).
    • Users can immediately drop/paste images and type prompts; if generation is requested before engine readiness, prompts are seamlessly queued and execute the moment the engine finishes booting.

Download:

Download ImageStudio.exe below and run!

Assets 3
Loading