Concurrent Submit | Zhenzhen_Flux_Kontext

This is the base Kontext node in the Comfyui-zhenzhen pack, and it's the most flexible of the three: it handles both text-to-image and instruction-based editing, takes a reference image, lets you dial guidance yourself (the edit nodes don't), outputs 1–4 images, and supports iterative re-edits via its clear_image flag. It's the one you'd reach for when you want Kontext's in-context editing with a bit more manual control - or when you're doing multi-image reference work and need the extra headroom.

Kontext is BFL's instruction editor - the KB's panel on it calls it the strongest character-consistency tool of the Flux 1 era, since it holds a subject across successive edits far better than adapter-based approaches. This node resells the API through the author's shop: no 12B weights on your disk, no VRAM, just a key.

How the concurrent bit works

The plain Comfly_Flux_Kontext node blocks until the job is done. The Submit variant runs it in the pack's shared image pool (30 workers) and returns a task handle instantly. Collect via Concurrent Collect Images (30): wire task_1 … and the collector waits for the batch, returning images in slot order plus a status string. failure_mode = fail-fast (default) or placeholder-on-error.

The README's own example for this node family is generating several images in one go - "supports setting the number of images (1–4) and multi-image input." The concurrent wrapper just multiplies that by your whole batch.

The inputs that matter

• prompt (required) - generation prompt or edit instruction.
• input_image - optional reference. Leave it empty for text-to-image; connect it for editing.
• model - flux-kontext-dev / -pro / -max.
• aspect_ratio - includes a Default option plus 21:9 → 9:21.
• guidance - 1–20, default 3.5. The knob the edit nodes don't expose.
• num_of_images - 1–4 per call.
• clear_image - default true. Set it false and the node can edit the image it previously generated (single-image mode), the documented iterative-edit flow.

Output: one task (COMFLY_IMAGE_FUTURE) → Collector → images.

Installing it

Ships in Comfyui-zhenzhen:

• ComfyUI Manager → Install Custom Nodes → search "Comfyui-zhenzhen" → install → restart.
• Or cd ComfyUI/custom_nodes && git clone https://github.com/T8mars/Comfyui-zhenzhen and restart.

No model downloads; dependencies are the standard stack. You need a paid token from the author's storefront (ai.t8star.org), in api_key.

Gotchas

num_of_images and clear_image=false both multiply how much you spend per run - check your batch before you commit to four variants of four jobs. And remember Kontext rides BFL's moderation rails; the anti-circumvention license drama that followed the open-weights release doesn't change what the API will and won't produce. Standard pack notes: proxy for the overseas endpoint from mainland China, retry on 500s.