最近两个月更新了不少新的绘图和视频模型：Seedance 2.5、Gemini Omni Flash、Seedream 5.0 Pro，还有刚更新的 Image 2.5；甚至连

所以群里小伙伴问得最多的一个问题就是：这些模型，到底在哪里可以一次都用上？

毕竟每个平台单独充值月费，算下来也不是小数目；自己搞 API，又有一定学习成本；至于那些价格特别便宜的中转站，又总让人担心哪天突然翻车跑路。哦，对了，还有大伙头疼的网络问题（这里咋就不细说，懂得都懂）

靠谱的综合平台，有的，之前也测评分享过；今天来了个新家伙：PixTV。AI

PixTV 是一个集图片生成、视频生成和剪辑于一体的综合创作平台，在国内网络环境下也可以直接使用，没那么多后顾之忧。它把不同类型的图片模型和视频模型放进同一个工作台，像 image2.5、Nanobanan都可以快速出图，Seedance、MiniMax H3 可以生成视频，再配合画布完成素材整理、修改和流程衔接，让创作者多一个相对省事的选择。

忘了，其实还有音频生成，目前配有MiniMax Music 3.0模型 。

该有的基本配齐，接下来试试它的水平高低；首页新增画布进来后，可以快速点中间模块开始创建，也可以在左侧按钮点➕号新建。

图片生成

东方水墨海报

使用了image2.5的模型，提示词源自X平台@zhidawang219555；这效果一看就是正宗的image那味了，毕竟我刚在G家测评完一轮。

Plain Text请严格以我上传的参考图为唯一主体参考来源。首先分析并提取参考图中最具有辨识度的视觉要素，包括主体身份与类型、面部或头部特征、发型、耳朵、角、帽子等头部轮廓、服装、配饰、道具、身体姿态与动作方向、最有辨识度的外轮廓、主体之间的关系、原图中具有代表性的色彩，以及能够体现主体身份的标志性细节。不要机械描摹参考图，也不要追求照片级写实复制。在保证主体一眼仍然能够被认出的前提下，对参考图进行高度概括、抽象化和东方艺术化重新创作，将主体转译为具有强烈中国东方美学气质的现代抽象手绘线条艺术。整体采用极致简约的毛笔线条与现代线描语言，通过少量但富有表现力的曲线重新概括主体结构。线条具有明显的手绘毛笔感，粗细自然变化，带有轻微飞白、墨迹干湿变化与手工笔触，大量使用修长、流畅、飘逸的艺术曲线。部分线条可以从主体结构自然延伸出去，形成类似书法运笔、衣带、水流、风、云气一般的长曲线，使画面产生明显的速度感、方向感和东方韵律。不要把所有结构完整画出来，允许通过断线、留白、几笔概括、局部缺失和视觉联想来表现主体。重点保留最具有识别度的轮廓、姿态和标志性元素，其余细节大胆舍弃。色彩控制在极少数量，以黑色线条作为主要视觉骨架，并根据参考图本身的代表性色彩，提取1至2种强调色进行艺术化使用，优先允许使用黑色、朱红色、墨绿色。颜色只作为局部点睛，不进行大面积填色，部分标志性服装、配饰、道具或动作轨迹可以使用红色或绿色长线进行强调。构图完整但极度克制，主体位于画面视觉中心区域，可以适当打破常规比例，对身体、衣物、毛发、尾巴、道具或动作轨迹进行拉长和曲线化处理，让主体像由几根正在流动的毛笔线条自然形成，而不是由封闭轮廓勾勒出来。整体融合东方抽象艺术、现代极简主义、中国书法线条、水墨速写、传统神话插画、现代编辑插画的视觉语言。画面加入一枚小型传统朱红色篆刻印章，印章文字根据参考图主体的身份、名字、昵称、角色特征或画面主题进行提炼，控制在2至4个汉字以内。如果没有明确文字信息，则不要擅自添加具体姓名，可使用与主体气质相符的简短东方意象文字。印章只是视觉点缀，尺寸较小，不抢主体。背景保持极简，使用细腻的米白色宣纸作为底色，具有非常轻微、自然、真实的纸张纤维纹理，不要出现复杂场景、建筑、景物或装饰性背景。画面四周必须保留大面积留白，尤其主体外围不要塞满元素，利用留白与少量曲线形成东方绘画中的虚实关系和呼吸感。最终效果应像从一张真实参考图中提取灵魂与轮廓，再用几根飘逸的中国毛笔长线重新画了一遍。强调主体辨识度必须保留，但避免照片复制；结构高度概括，但不能抽象到无法识别；线条少而精准，飘逸但不凌乱；传统东方气质与现代设计感同时存在。高清细腻，艺术海报级完成度，干净、克制、高级。禁止写实照片效果、3D渲染、卡通Q版、复杂背景、大量颜色、大面积色块、密集细节、机械描边、封闭式矢量轮廓、霓虹灯效果、赛博朋克元素、多余文字、英文文字、商业Logo、水印。画面重点始终是参考图主体辨识度、极简毛笔线条、飘逸长曲线、东方留白、少量朱红或墨绿点睛。

首页新增画布进来后，在左侧按钮栏，点 + 就可以打开是生成图片，还是视频或音频等。中间是效果显示区，右边是编辑选择模型、尺寸、画质以及视频时长等选项。

活动宣传海报

使用的Seedream5.0pro模型，效果是最好的。

Plain Text节庆活动海报，3:4竖版构图，亮黄撞色主背景搭配浅蓝色细边框。画面正中是一只巨大的绿色毛绒怪物角色，圆眼睛，表情慵懒松弛，毛绒质感明显，体积感柔软饱满。怪物周围错落点缀星星、笑脸、彩色小徽章、趣味图形和夸张装饰字体，整体视觉可爱怪诞、活泼跳脱、年轻有趣，具有潮流节庆活动主视觉气质，契合深圳城市青年文化与生活方式活动氛围。构图强调中心角色与文字之间的互动关系，装饰元素丰富但不过度拥挤，整体具有高完成度的商业活动海报设计感。文字排版要求清晰准确：海报顶部使用粉色立体英文大字「Lazy Bloom Fest」，沿弧形向上弯曲排布；围绕怪物身体中部加入白色弧形文字「SOUTH COAST CREATURE FAIR」；海报底部设置醒目粗体主标题「2026 怪趣生活节」；主标题下方补充活动信息「2026.09.17—09.25｜深圳・海岸城广场」；最底部以较小字号标注「装置展 / 夜市 / 怪物巡游 / 限定周边」。所有中英文文字必须完整、清晰、无错字，不增加无关小字、水印或多余品牌信息。

节日黑板报

我猜老师们应该要开始筹备双节黑板报了，这次做得是初中版，方便先画出来，再粉笔复刻。这里对比后，还是image发挥得最好。

Plain Text图1：横版初中教室黑板报，深绿色黑板底，粉笔手绘质感，中秋国庆双节同庆主题，融入深圳湾区城市特色。对称式新中式布局，顶部居中是横版主标题 “湾区明月・家国团圆”，字体端庄大气，朱砂红粉笔字搭配桂花黄描边。标题下方用一条飘逸的红色长飘带连接左右版面，飘带中写 “庆国庆”“迎中秋”。左侧绘制简化华表、红灯笼和五角星，右侧绘制满月、垂落桂花枝和玉兔，底部绘制一条极简深圳湾区天际线剪影，搭配浅蓝海浪线。中间设置三块整齐文字区，标题分别写 “祖国颂”“团圆月”“深圳情”，正文区域用整齐横线表示可书写内容。图2：横版初中教室黑板报，深绿色黑板底，粉笔手绘质感，中秋国庆双节同庆主题，融入深圳海滨城市特色。中心式构图，画面正中央是一个巨大的圆形月饼主视觉，饼面采用左右双拼设计：左半面绘制五星红旗五星元素，右半面绘制简笔月兔望月图案，巧妙融合双节主题。月饼四周均匀分布 4 个圆角文字框，框内书写工整的简短节日文字。版面空白处错落点缀简笔画海浪、小帆船、椰树、迷你月饼和小灯笼装饰。顶部主标题「月饼双拼・双节同乐」为圆润可爱的艺术字，外圈带有月饼压花纹边框。整体线条简洁流畅，全部采用块面平涂无渐变，配色以橙红、暖黄、浅天蓝为主，色调活泼明亮，绘制难度低，适合学生复刻，版面干净饱满不杂乱。

还有审美依然领先的Midjourney

试了最近很热门的宋韵少女风格系列~有种电影感feel

Plain Text溪边拾叶拨水：白色与翡翠绿宋制汉服 ✨ 溪边俯身拾起落叶 ✨ 指尖轻触水面，荡开清澈涟漪 ✨ 前景虚化、逆光光斑、温柔电影感A young woman in simple white and emerald Song-style Hanfu kneels gracefully by a clear riverside in early autumn, leaning forward to pick up a fallen leaf while lightly brushing the water with her fingertips. Focus on her gentle lively movement, soft smiling expression, elegant body curve, tiny sunlit droplets, crystal-clear ripples, smooth river stones, scattered autumn leaves, vibrant green riverside grass in the softly blurred foreground, distant glowing foliage, cinematic natural light, shallow depth of field, realistic photography, warm highlights, detailed fabric texture and refined water texture.庭院晾衣：白色与翡翠绿宋制汉服 ✨ 秋日小院晾晒衣物 ✨ 手提湿衣，衣角滴水，动作自然 ✨ 柔和阳光、前景虚化、真实电影感A young woman in simple white and emerald Song-style Hanfu stands in a quiet autumn courtyard, hanging freshly washed clothes to dry on a bamboo pole or wooden clothes rack. Focus on her graceful natural movement, soft joyful expression, elegant posture, and flowing Hanfu sleeves as she lifts the damp fabric with both hands. Water droplets fall lightly from the hem, catching warm sunlight. The courtyard features stone paving, scattered fallen leaves, a low wall or wooden fence, potted plants, bamboo baskets, softly blurred foreground branches, sunlit foliage in the background, shallow depth of field, cinematic natural light, realistic photography, warm highlights, detailed fabric and wet cloth texture, and a subtle poetic everyday atmosphere.

Plain Text1.Moebius (Jean Giraud)风格，极繁主义，极致表现力，大师级视角，浪漫感，细节完美，大师杰作。秋日森林秘境，梦幻童话世界，画面干净通透，色彩丰富而富有层次。森林进入一年中最绚烂的秋季，树木呈现明亮鲜艳的金黄色、橙红色、暖棕色、浅绿色与少量清透蓝色，大片金黄树冠彼此交错，落叶轻盈飘落，林间小径铺满彩色秋叶，阳光从枝叶缝隙穿透形成柔和斑驳光影。加入蜿蜒溪流、奇异植物、蘑菇、小花、岩石、藤蔓与远处起伏的梦幻山丘，局部点缀童话般的小型生物与奇异自然景观，使画面充满探索感与惊喜感。整体氛围奇幻、可爱、温馨、浪漫，带有清爽明亮的秋日空气感与丰收季节的生命力，避免萧瑟、阴冷和灰暗感，仿佛进入一个被金色阳光包裹的秋日绿野仙踪。autumn forest, golden autumn, colorful fall foliage, falling leaves, warm sunlight, dreamy woodland, magical landscape, whimsical fantasy, golden yellow, orange red, warm brown, fresh green, clear blue sky, soft breeze, luminous atmosphere2.hand-drawn illustration style, soft pastel palette, a cheerful little girl laughing under sunlight, wearing a bright yellow sweater and wide hat, wavy hair gently flowing, holding a drink cup, bright pure smile, background with flowers and abstract color splashes, warm and airy atmosphere, watercolor and colored pencil texture, pastel brushstrokes, visible paper grain, gentle light bloom, healing and dreamy Japanese illustration, high-quality hand-drawn aesthetic, V7

图片编辑：点击图片，上方有一排图片编辑的按钮，这个跟多数平台差不多，按需选择即可。

视频生成

MinimaxH3模型

[7644d742e148470e9e272da3afd5aa22.mp4]

提示词源自X平台@ImaStudio_ai

上面15s视频才花了150积分，H3生成视频真的很划算；所以我忍不住继续来一份。

设计好品牌图片后，后续的视频提示词也可以直接在 PixTV 里完成。点击左侧新增文本，在【文本增强】功能中，就能调用最新的 GPT-6 Astra，同时也已经配置了 Claude、Gemini 等主流模型，用来润色提示词、补充镜头细节或者优化视频描述都很方便。

还有一个很实用的优势，就是不受网络环境限制，国内可以直接使用。

图片提示词改写

视频提示词输出

[bf0d6c1b76c2411a8d862edc8c288b31 (1).mp4]

SQLsubject_definitions: <Subject 1> is the complete “桂月茶点局” mooncake gift-box system consistently shown across <Picture 1> to <Picture 7>: a low-saturation grayish celadon-green rigid square box, large warm-gold foil moon, right-side osmanthus branch, restrained mountain and cloud patterns, floral-window emblem, and the exact visible Chinese brand name. <Subject 2> is the same round osmanthus oolong lava mooncake: floral scalloped edge, refined Chinese embossing, natural golden-brown surface, thin pastry skin, fine gray-brown oolong filling, and a restrained amber-gold osmanthus lava center. <Subject 3> is the complete six-piece set: exactly six individually wrapped mooncakes arranged in a two-by-three grid inside the same light-khaki tray. <Subject 4> is a warm adult female Mandarin voice-over, calm, intimate, refined, and natural. <Picture 1> provides the opening hero composition. <Picture 2> provides the box texture, gold foil, printing, and construction details. <Picture 3> provides the exact six-piece quantity, wrapping, tray, and arrangement. <Picture 4> provides the mooncake cross-section and lava texture. <Picture 5> provides the osmanthus, oolong tea, porcelain ware, and flavor composition. <Picture 6> provides the gift-giving hand interaction and Chinese interior. <Picture 7> provides the final brand packshot and closing composition. summary: [reference generation + keyframe completion] Create a 15-second, 9:16 premium Mid-Autumn e-commerce commercial for “桂月茶点局”. Use <Picture 1> as the opening visual and <Picture 7> as the final visual. Present the product through seven concise shots: hero reveal, packaging craftsmanship, six-piece box, lava filling, osmanthus oolong flavor, gift-giving, and brand closure. retention_analysis: Keep <Subject 1> identical in every shot. Preserve the celadon-green color, rigid square structure, proportions, gold-foil moon, osmanthus branch, mountain and cloud patterns, emblem, and visible Chinese brand name. Keep <Subject 2> consistent in shape, embossing, baked color, filling layers, and amber-gold lava texture. Keep <Subject 3> as exactly six mooncakes in a two-by-three grid. Never add, remove, duplicate, merge, or rearrange them. Maintain one unified commercial photography language: warm natural side light, dark walnut wood, off-white porcelain, beige linen, celadon green, warm gold, and small osmanthus-yellow accents. Use clean editorial cuts instead of product morphing. Keep package edges straight, hands natural, food realistic, and brand text stable. No extra logos, promotional text, unrelated products, red festive backgrounds, or exaggerated effects.detailed_description: [Shot 1] Premium 9:16 opening based on <Picture 1>. The closed gift box stands behind the open box, with exactly six individually wrapped mooncakes visible. Warm side light moves gently across the gold-foil moon. The camera performs a slow, restrained push-in. Osmanthus flowers move slightly. <Subject 4> narrates: <d>[ZH]桂花入月，乌龙入心。</d> [Shot 2] At 00:02.200, cut to the macro view from <Picture 2>. The camera slides slowly across the celadon paper texture, embossed gold moon, osmanthus branch, box edge, and brand name. A narrow highlight reveals the foil and printing depth. The package remains rigid and unchanged. [Shot 3] At 00:04.000, cut to <Picture 3>. Begin close on one transparent wrapper, then gently crane upward to reveal the complete two-by-three arrangement. Exactly six wrapped mooncakes remain visible. Soft reflections move across the transparent film while the box and tray stay stable. [Shot 4] At 00:06.000, cut to <Picture 4>. Use a shallow-depth-of-field macro shot. Rack focus from the mooncake surface pattern to the inner pastry, oolong filling, and amber-gold lava center. The thick lava moves downward slowly and naturally without exaggerated bursting. A few crumbs settle on the ceramic plate. <Subject 4> narrates: <d>[ZH]一口流心，藏住整个秋天。</d> [Shot 5] At 00:08.500, cut to <Picture 5>. The camera makes a gentle semicircular move around the cut mooncake, osmanthus, oolong leaves, porcelain gaiwan, and amber tea. Fine steam rises from the tea and several blossoms move subtly. The same gift box remains recognizable in the softly focused background. [Shot 6] At 00:10.800, cut to <Picture 6>. Two pairs of natural adult hands complete one slow gift-box handover. The box stays front-facing, rigid, readable, and unchanged. Soft round-window light falls across the hands and gold-foil moon. The gesture feels warm and sincere. [Shot 7] At 00:12.800, cut to <Picture 7>. The closed gift box and one cut lava mooncake sit against generous warm off-white and pale-celadon negative space. The camera slowly pulls back and becomes still. A soft moon-shaped shadow and faint osmanthus shadow settle behind the product. Keep the brand name sharp until the final frame. <Subject 4> concludes: <d>[ZH]桂月茶点局，月满，心也圆。</d> overall_soundscape: Native stereo sound with subtle box movement, soft wrapper rustle, gentle porcelain contact, quiet tea pouring, and a faint breeze through osmanthus leaves. Keep the Mandarin voice-over warm, close, clean, and clearly separated from the background. non_diegetic_music: Refined contemporary Chinese instrumental music at about 76 BPM, using soft guqin and ruan plucks, warm ambient pads, light wooden percussion, and sparse bell accents. Begin intimate, rise gently during the six-piece reveal and lava close-up, then resolve into one warm sustained note on the final brand frame.

Seedance2.5模型

2.5模型出的视频，人物皮肤细节真的太对人味了，很真实，我自认垫图都没那么好。

[63ca7bba957f4fc08952b72f0d194b7a.mp4]

提示词源自X平台@BubbleBrain。

SQLCreate a 15-second Japanese-style idol personal vlog following the same young performer through a compressed private-day diary. Keep her face, hairstyle, wardrobe logic and personality fully consistent. The footage should feel self-shot or casually filmed by a close friend, with handheld phone framing, slight focus misses, ordinary room light, natural pauses, playful reactions and clean native ambience. Structure the 15 seconds as a fast but natural diary progression: 0–3s, direct-to-camera morning greeting at home, sleepy and casual; 3–6s, quick breakfast or getting-ready details, then a candid moment playing with her fluffy cat; 6–9s, handheld travel transition and a brief convenience-store or cafe food stop;9–12s, rehearsal-room preparation with one short energetic performance fragment; 12–15s, an unguarded backstage smile or tired reaction, then a quiet nighttime sign-off to camera. Use seamless match cuts and natural handheld transitions so the day feels continuous rather than like unrelated clips. Let her occasionally speak short natural Japanese phrases directly to camera, with believable lip sync, changing room tone and realistic acoustic shifts between home, street, cafe and rehearsal room. Avoid subtitles, on-screen text, logos, commercial posing, glamour-ad lighting, face drift, duplicate subject, plastic skin, distorted hands, floating props or overly polished cinematic staging.

Gemini Omni Flash

提前生成LOGO刺绣的图片，然后垫图用Omni生成。

SQLExtreme macro shot of woven cotton shirt fabric, empty at first. Colored embroidery threads rise from the surface and weave themselves into place, laying down tight satin stitches row by row, tracing out the logo. The stitched area grows steadily across the frame, thread pulling taut and raising into dimensional embroidered texture, until the full logo is complete. Camera holds steady, slow satisfying timelapse pace. Soft natural side light, shallow depth of field, visible fabric fibers and realistic thread sheen. No hands, thread moves on its own. One shot no scene change.

[视频剪辑-edited.mp4]

这里我是生成了多个LOGO的刺绣视频，直接用PixTV的视频剪辑把多段视频剪辑成一个视频。

视频剪辑：多段视频组合可进行分割、音频分离、变速等基础剪辑功能；可以选择下载本地或导出画布再次处理。

好了，玩了一个周末，PixTV的测评分享就到这里。

这一圈测下来，PixTV 给我最直观的感受就是，文本、图片、视频到后期剪辑基本都配齐了，一套平台就能完成大部分创作流程，不用在几个工具之间来回切换，也顺手解决了最让人头疼的网络问题。实际测评体验下来，主流模型的生成效果也基本在线，出图、出视频的稳定性和完成度都不错。对于不想折腾 API，又希望一次体验多个新模型的人来说，确实省事很多，尤其适合日常高频做图、做视频和内容创作的人。