zhenzhen-sensenova-u1-infographic-fal

Most image models can't do text - you ask for a chart and you get noise squiggles that mean nothing. SenseNova U1 is China Mobile's answer to that specific problem: an infographic model that lays out data, labels, and charts into something that looks designed. Comfly_sensenova_u1_infographic_fal is the zhenzhen pack's FAL route to it. If your workflow occasionally needs to produce a shareable explainer graphic - and you're tired of faking it in an image editor - this is the node.

It's a plain FAL text-to-image: prompt in, generated infographic out, through zhenzhen's proxy at https://ai.t8star.org/fal. The token goes in the node's api_key field, the usual submit-and-poll flow applies, and prepaid credits get settled after the job. The fun part is that this one exposes actual sampler-style knobs, which is rare for an API node.

The inputs that matter

• prompt - required. This is where you describe the infographic: "A 3-step process diagram explaining how solar panels work, with labeled arrows and a small pie chart." Structure talk is what this model understands.
• aspect_ratio - 1:1 through 21:9-style presets, 16:9 default.
• guidance_scale - 0–20, default 4.0. How hard it follows the prompt.
• num_inference_steps - 1–100, default 50. The sampler dial.
• timestep_shift - 0–10, default 3.0; it biases which part of the denoise schedule does the work.
• use_thinking - off by default. When on, the model reasons about the layout before drawing, which helps complex infographics but costs time and credits.
• sync_mode - off by default; on for synchronous (non-queue) execution. Leave off unless the queue is misbehaving.
• seed - 0 = random, cap 65535 (the fal limit).

Outputs: images (tensor), image_urls, and response. Poll defaults are the pack's standard 6s × 600.

Installing it

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/Comfyui-zhenzhen

or ComfyUI Manager (search "Comfyui-zhenzhen"), restart, paste a zhenzhen token into api_key. No models to download.

Common issues

• Garbled text in the output - infographic models still garble small labels. Ask for fewer, bigger labels and check the render before using it anywhere important.
• Flat or unreadable layout - raise guidance_scale a bit and consider use_thinking for multi-part layouts. Re-roll; this model's variance is real.
• Cost creep - higher steps + thinking + big aspect ratio all push the fal pre-charge up. The README notes fal pre-charges and refunds the difference, but plan for the worst case.

For a beginner the honest advice is: treat this as "infographic generator that's already better than a diffusion model faking a chart." It won't replace a real data-viz tool for precision work, but for a shareable explainer it's surprisingly good - and it's the only node in this pack that lets you tune guidance scale.