SeeDream 5.0 Lite (BudgetPixel)

Of the three Seedream image options in the BudgetPixel pack, 5.0 Lite is the one you'll actually reach for on a Tuesday. It's the fast, cheap tier of ByteDance's Seedream 5.0 line: 40 credits per generation against Pro's 55, unified text-to-image and editing in one node, up to nine reference images for multi-reference composition, and 2K/3K output. It's not the top-quality tier - that's Pro - but for anything that isn't a hero shot, the gap is smaller than the price gap, and it's the Seedream you want when you're iterating on prompts and watching the credit counter.

What it does differently

Same hosted-model shape as every BudgetPixel node: connect images or not, type a prompt, and the job runs on BudgetPixel's servers while your local GPU idles. The interesting bits are the grouped-generation mode and the price. Set sequential_image_generation to auto and the model returns a set of related images rather than a single one - up to max_images, which goes to 14 here (Pro doesn't offer grouped generation at all). The model picks the exact count and you're billed per image returned, so max_images is a ceiling you set, not a number you get.

The inputs that matter

• prompt - required. Text-to-image with no references; editing/composition with them.
• reference_images - up to 9. Multi-reference composition is the reason to pick this over a plain text-to-image service.
• size - 2K (default) or 3K.
• num_images - 1–4 for plain batch generation (billed per image).
• sequential_image_generation / max_images - the grouped mode described above; leave at disabled if you want exactly one image and no surprises on the bill.

Output is images (IMAGE), ready for Save Image or whatever's downstream.

Install and key setup

Everything is the shared BudgetPixel story: ComfyUI Manager → search BudgetPixel → install and restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/budgetpixel-ai/comfyui-budgetpixel.git
pip install -r comfyui-budgetpixel/requirements.txt

Only dependency is requests - no model downloads, no GPU needed. API key from a paid plan via BUDGETPIXEL_API_KEY or a budgetpixel.json file in the pack folder; your key never lands in workflow JSON, so shared graphs stay clean.

Where people get burned

The "lite" tier is a cost move, not a quality promise - if you need precise layer-level editing, Pro is the node with that control and Lite doesn't pretend to have it. Grouped generation bills per returned image, so watch max_images. And the pack-wide pair of traps apply as always: ComfyUI caches unchanged outputs (hit force_rerun to force a fresh, billed generation), and free accounts get api_access_not_enabled until you're on a paid plan that includes the API.