Flux 2 Klein 9B_txt2img - LM Studio Prompt Enhancer - ornith-1.0-9b-mtp

This workflow requires my Comfyui-lmstudio-prompt-enhancer custom node which is installable via Comfy Manager.

🛈 IMPORTANT!!!
You will need to search for, download, and load the model ornith-1.0-9b-mtp in LM studio. You will also need to start the API server and generate an API Key. Learn how to configure LM Studio here.

My rating of ornith-1.0-9b-mtp Flux prompt generation ability is: 8.0/10

Prompts critiqued: 12
Average score: 8.0 / 10
Highest score: 8.6 / 10
Lowest score: 7.7 / 10

A critique of ornith-1.0-9b-mtp Flux prompt generations abilities:

The LLM ornith-1.0-9b-mtp shows a strong instinct for cinematic image construction and clearly understands the kinds of visual cues Flux responds to: lighting direction, material detail, atmosphere, camera angle, and subject scale. Their prompts consistently create compelling scenes with strong mood, especially through color palettes, reflections, fog, rain, glowing energy, and texture descriptions. The main weakness is a tendency toward prompt density and over-specification, where too many effects, style labels, or precision demands compete for control. The author often asks Flux to preserve exact logos, perfect reflections, strict vehicle fidelity, or complex optical behavior, which reduces reliability even when the concept is strong. Their best prompts are the simpler, more grounded ones with a clear subject hierarchy and concrete visual details. Overall, the author is above average and capable of producing strong Flux prompts, but would improve further by trimming decorative language, reducing fragile requirements, and giving the main subject more specific design control.

Notes for using ornith-1.0-9b-mtp

• This model this model is fast I got on average 44 tokens per second on my AMD Strix Halo RyzenAI Workstation. -generates detailed prompts, with minimal rambling. The generated prompts produces images with expert composition and color grading.

• I set the max tokens to 2500, this higher value is to accommodate this model's reasoning.

• If the image looks flat, check the logs in LM Studio, the last log entry will contain a JSON path x.choices[0].message.content if the content field is empty that means the LLM ran out of reasoning tokens, this can be resolved by increasing max_tokens.

• this model tends to start the enhanced positive prompt's string with newline characters "\n" it does not appear to cause any problems.