The Hy Image 3.5 prompt generator gives you 20 free, copy-paste prompts for Tencent Hunyuan's image model. Sharper text in images, realistic portraits, product shots and reference-image edits, each ready to paste and adapt.
This Hy Image 3.5 prompt generator collects 20 tested-style prompts for Hy Image 3.5 Preview, Tencent Hunyuan's image generation and editing model. The model handles text-to-image, image-to-image and multi-round editing, and Tencent says it improves on Hy Image 3.0 in text rendering, realism, style and editing consistency.
It is available through Tencent Cloud's TokenHub, OpenRouter and ComfyUI partner nodes. Copy any prompt below, swap in your own subject, and run it on the platform you use.
Prompts 10–14 need one or more attached reference images
A glowing neon sign on a dark brick wall that reads exactly 'OPEN LATE' in curved pink and blue tubing, subtle reflections on a wet pavement below, shallow depth of field, photorealistic night photography.
A hand-lettered chalkboard menu in a small cafe listing 'Flat White — 4.50', 'Cold Brew — 5.00' and 'Almond Croissant — 3.75' in neat white chalk, warm window light from the left, a ceramic cup on the counter in the foreground.
A cinematic movie poster for a sci-fi thriller titled 'LAST SIGNAL' in bold condensed metallic letters at the bottom, a lone astronaut silhouetted against a huge red moon, teal and orange colour grade, tagline in small type above the title.
A realistic portrait of a woman in her thirties by a window, soft side light, natural skin texture with fine pores, slightly tousled dark hair, grey knit sweater, calm direct gaze, 85mm lens look, muted background.
A candid street photograph of an elderly man laughing at an outdoor market stall, motion in the background, golden hour light, 35mm documentary style, shallow depth of field, authentic colour.
A minimalist product photo of a frosted glass perfume bottle on white marble, soft diffused light, a single green leaf casting a delicate shadow, clean negative space on the right for text, high-end advertising style.
A white running sneaker floating above a pastel peach background with soft shadow beneath, crisp stitching and mesh texture visible, studio lighting, commercial e-commerce hero image.
A top-down flat lay of a rustic breakfast: sourdough toast with avocado and poached egg, a glass of orange juice, linen napkin, wooden table, natural morning light, appetising colour and fine detail.
A cosy Scandinavian living room with a cream bouclé sofa, a wool throw, a low oak table and a large leafy plant, soft afternoon light through sheer curtains, interior design magazine photography, wide angle.
Using the attached reference image, keep the same character, face, hairstyle and outfit exactly, and place them walking through a rainy neon-lit Tokyo alley at night, cinematic lighting, holding a transparent umbrella.
Using the attached photo of the person and the attached photo of the jacket, dress the person in that jacket while keeping their face, pose and background unchanged, natural fabric folds and consistent lighting.
Combine the people from the attached reference images into one friendly group photo at a picnic in a park, each person keeping their own face and clothing, same afternoon lighting for everyone, natural spacing.
Edit the attached image: replace the background with a sunlit mountain meadow, keep the subject exactly as is with clean edges, match the lighting direction and colour temperature of the new background.
Change the attached photo from midday to blue-hour dusk: deepen the sky, switch on warm window lights in the buildings, keep all structures and people unchanged, subtle long-exposure feel.
An anime-style illustration of a girl with silver hair and a red scarf standing on a snowy rooftop at night, soft rim light from city lanterns, clean line art, painterly background, high detail.
A loose watercolour painting of a fishing village at sunrise, bleeding washes of coral and teal, visible paper texture, small boats in the harbour, delicate ink outlines.
An isometric 3D render of a tiny bedroom diorama with a desk, bed, bookshelf and glowing lamp, soft clay material, pastel palette, tilt-shift look, clean studio background.
A clean infographic poster titled 'DAILY WATER INTAKE' with five simple icons in a row, short labels beneath each icon, a blue and white palette and generous spacing, flat vector style, all text sharp and legible.
A fashion editorial photo of a model in an oversized camel coat on an empty concrete staircase, hard directional sunlight, long shadows, minimalist composition, medium format film look.
A modern concrete house with floor-to-ceiling glass in a pine forest at dusk, interior lights glowing warm, a still reflecting pond in front, architectural photography, wide shot, rich detail.
| Model | Maker | Strengths | Prompts |
|---|---|---|---|
| Hy Image 3.5 Preview ★ | Tencent Hunyuan | Text rendering, multi-reference editing | This page |
| Seedream 5 Pro | ByteDance | Photorealism, consistency | Seedream 5 Pro prompts |
| GPT Image 2 | OpenAI | Conversational editing, broad skill | GPT Image 2 prompts |
| Nano Banana 2 | Fast, photo-real edits | Nano Banana 2 prompts | |
| Qwen Image 3 | Alibaba | Text and layout, open ecosystem | Qwen Image 3 prompts |
| Reve 2 | Reve | Layout-first composition | Reve 2 prompts |
Capabilities summarised from vendor announcements for Hy Image 3.5 Preview; reference-image limits and maximum resolution are reported differently by different sources.
The Hy Image 3.5 prompt generator on this page gives you 20 free, copy-paste prompts for Hy Image 3.5 Preview, the image generation and editing model from Tencent Hunyuan. The prompts cover text rendering, realistic portraits, product shots, reference-image edits, illustration and design, so you can see what the model does well and adapt each prompt to your own idea.
Hy Image 3.5 Preview is Tencent Hunyuan's image model, released in preview in late September 2026. It supports text-to-image, image-to-image and multi-round editing, and Tencent says it improves on Hy Image 3.0 in text rendering, realism, style range and editing consistency. Details such as the number of reference images and maximum resolution are reported differently by different sources, so check the platform you use for its current limits.
Hy Image 3.5 Preview is available through Tencent Cloud's TokenHub platform, through OpenRouter, and through partner nodes in ComfyUI. Tencent is also integrating it into its own products such as the Yuanbao chatbot. Availability and pricing change quickly during a preview, so confirm on the provider's page.
Describe the subject, the setting, the lighting and the camera or art style in plain sentences. When you need text inside the image, put the exact wording in quotes and say where it appears and in what style. For edits, say clearly what must stay unchanged (face, pose, structure) and what should change. Short, specific prompts usually beat long lists of adjectives.
Yes. It accepts reference images for consistent characters, outfit changes and multi-person scenes. The supported number of reference images varies by platform and is reported differently across sources, so start with two or three references and add more only if your platform allows it. Prompts 10 to 12 on this page are written for reference-image workflows.
Tencent reports that in an internal blind test Hy Image 3.5 Preview scored on par with Seedream 5.0 Pro. That is a vendor result, not an independent benchmark, so test the same prompt in each tool. See the comparison table above and our pages for Seedream 5 Pro, GPT Image 2 and Nano Banana 2 to compare prompting styles.