Six models × 1400+ bilingual lexicon

Sdxl Prompt Builder

All-in-one AI drawing prompt builder. It supports six major models: Illustrious, Pony, Anima, and FLUX. It has a foldable vocabulary, weighted grammar, and parameter suggestions to accurately control the generation details.

SDXL Universal sampler DPM++ 2M Karras number of steps 25-30 CFG 5-7 Clip Skip 1-2 Resolution 1024×1024

What is an SDXL prompt word generator?

Sdxl Prompt Builder It is a free AI drawing Prompt construction tool. It has more than 900 sets of Chinese and English bilingual lexicon built-in, allowing you to quickly assemble professional-level drawings by clicking on the tags. Stable Diffusion Prompt Word. Whether using SDXL、Illustrious XL、NoobAI、 Pony Diffusion、Animagine XL、Anima Still FLUX, can combine characters, clothing, poses, scenes, lights and styles with one click, and automatically generate English label output that conforms to A1111 / ComfyUI syntax.

How to quickly create a complete set of prompts?

Six model prompt word strategies

SDXL Universal General use

The basic SDXL model is suitable for all scenarios. Quality starts with:masterpiece, best quality, highres, ultra-detailed, absurdres, sharp focus. Recommended parameters: DPM++ 2M Karras, steps 25–30, CFG 5–7, resolution 1024×1024. Negative direction must be added bad hands, extra digits Wait for hand tags.

Illustrious XL / NoobAI anime

Currently the most mainstream anime model series in the community, it has an excellent understanding of the Danbooru tag. Quality starts with:masterpiece, best quality, amazing quality, very aesthetic, absurdres, newest. It is recommended that quality words be placed in prompt wordsending, if the scene is complex, you can put them once at the beginning and at the end. Recommended parameters: DPM++ 2M Karras or SGM Uniform, steps 25–28, CFG 5–5.8.

Pony Diffusion Rating word family

The Pony series features score_1 Arrive score_9 Rating label training, the positive direction must be written score_9, score_8_up, negative join score_4, score_5, score_6 Exclude low quality. Recommended parameters: Euler A, steps 28–40, CFG 7–9, Clip Skip 2.

Animagine XL anime

Animagine XL 4.0 series animation model supports year labels (year 2024) and the style tag. Quality starts with: masterpiece, best quality, highly detailed, anime style, beautiful detailed face. Recommended parameters: DPM++ 2M Karras, steps 28–30, CFG 5–7, Clip Skip 2.

Anima Next generation 2B parameters

Anima (CircleStone Labs and Comfy Org) is a new generation animation model that accepts both Danbooru tags and natural language descriptions. Official recommended starting moves:masterpiece, best quality, score_7, safe, and withLowercase, space separated Tags (except score tag). Be careful to avoid excessive stacking of image quality words. Rich content tags are more effective than negative stacking. Recommended parameters: er_sde, number of steps 30–50, CFG 4–5, Clip Skip 1.

FLUX/Natural Language Model natural language

FLUX, Qwen-Image and other new generation model preferencesnatural language description rather than pure labels. You can concatenate the selected tags into complete sentences, for example:a stunning anime masterpiece of a silver-haired knight, highly detailed, 8k, sharp focus, cinematic lighting. FLUX is less sensitive to negative cue words and requires only a small number of quality words to be negative.

Prompt word weight grammar teaching

Weight is the core syntax for controlling the focus of generation. After turning on the "weighted mode" in this tool, clicking on the labels will cycle through them in sequence. 0.8 → 1.2 → 1.5 → 2.0, the output looks like (masterpiece:1.2) A1111/ComfyUI compatible syntax. The higher the weight, the stronger the element’s presence in the picture; 0.8 The following is used to downplay unimportant elements. Recommended for key characters or core styles 1.2–1.5, to avoid excessive weight leading to unbalanced composition.

The Importance of Negative Prompts

negative cue words Describing the elements you don't want to appear is key to stable build quality. Common negative elements include: Image quality issues (lowres, jpeg artifacts, blurry), anatomical errors (bad anatomy, bad proportions), hand defects (bad hands, extra digits, missing fingers), redundant elements (watermark, signature, text) etc. This tool has pre-prepared "recommended negative directions" for each model, which can be applied with one click after switching models.

Advanced techniques to improve image quality

Frequently Asked Questions (FAQ)

What AI drawing models does the SDXL prompt word generator support?
Supports six mainstream models:SDXL Universal, Illustrious XL / NoobAI, Pony Diffusion, Animagine XL, Anima (Anima-xz series) and FLUX / Natural Language Model. After switching the model, the quality starting position, suggested negative direction, and generated parameter suggestions will be automatically switched.
How to write the quality words of Illustrious XL or NoobAI?
Recommended masterpiece, best quality, amazing quality, very aesthetic, absurdres, newest. It is recommended that the quality word be placed at the end of the prompt word. When the scene is complex, it can be placed once at the beginning and once at the end.
Why does Pony Diffusion use the score_9 scoring word?
The Pony series features score_1 Arrive score_9 Rating label training,score_9 Represents the highest aesthetic quality. write forward score_9, score_8_up, and join in the negative direction score_4, score_5, score_6 to avoid low-quality results.
What is the difference between Anima model prompt words and SDXL?
Anima accepts both Danbooru tags and natural language descriptions, recommendedlowercase + space Separate tags and avoid excessive stacking of quality words. Official recommended starting moves:masterpiece, best quality, score_7, safe, CFG 4–5, steps 30–50.
Negative Prompt What is the negative prompt word? Why is it important?
Negative cue words describe elements you don't want to appear, e.g. lowres, bad anatomy, bad hands, watermark, text Wait. It can effectively reduce common defects such as deformed fingers, duplicate objects, and watermarks, which is the key to stable production quality.
How to use the prompt word weight Weight?
After turning on the "weighted mode" of this tool, click on the vocabulary label to cycle through it in sequence 0.8 → 1.2 → 1.5 → 2.0 weight, output (標籤:權重) A1111 compatible syntax. The higher the weight, the more emphasis is placed on the element.
How to set up CFG Scale, Steps and Sampler?
Each model recommends different values: Illustrious suggests CFG 5–5.8, steps 25–28, DPM++ 2M Karras; Pony suggests CFG 7–9, steps 28–40, Euler A; Anima suggests CFG 4–5, steps 30–50, er_sde. The tool will automatically display corresponding suggestions when switching models.
What should I do if there is a finger or deformation problem in the generated image?
First confirm that the negative direction is included bad hands, extra digits, missing fingers, bad anatomy; Then check if there are too many quality words stacked in the forward direction; Finally, you can increase the CFG or number of steps, and switch to a more stable sampler such as DPM++ 2M Karras.
Copied