ego6

2.3K posts

ego6 banner
ego6

ego6

@kidemc

- import export - noire & noich mal entourée .

เข้าร่วม Şubat 2012
4.7K กำลังติดตาม131 ผู้ติดตาม
ego6 รีทวีตแล้ว
Viola Zhou
Viola Zhou@violazhouyi·
How AI short dramas are currently made in China: young workers, earning some $500 a month, working at 3am to generate videos on Seedance 2.0 — because the model is cheapest, fastest after midnight. Report by 36kr (in Chinese) mp.weixin.qq.com/s/FLKHPnu2bl_R…
English
5
19
211
15.5K
ego6 รีทวีตแล้ว
ego6 รีทวีตแล้ว
阿绎 AYi
阿绎 AYi@AYi_AInotes·
holy shit, 微软直接把之前因为太危险被下架的语音AI王炸,给重新开源了, 这个叫VibeVoice的项目,GitHub已经冲到了29.3k星, 做播客的,搞有声书的,做AI视频配音的,还有天天要转写会议记录的兄弟, 谁没被付费语音工具的订阅费坑过,谁没被长音频处理崩过, 谁没被本地部署的复杂程度搞到头疼, 这个项目直接把这些痛点全给解决了, 核心能力直接爆表: - TTS:一次生成90分钟多说话人(最多4人)对话,情感、自然转接、长上下文超强 - 实时流式版(0.5B模型):延迟仅~300ms - ASR:单次处理60分钟音频,自动说话人分离 + 时间戳 + 50+语言 完全本地运行,免费无订阅,还带了不可感知水印 + AI声明防滥用。 2025年因为deepfake风险被Microsoft自己下架过,现在加了安全措施重新上线,真正的“核弹级”开源项目。 这下ElevenLabs要慌了😂 github 地址评论区自取👇 #VibeVoice #开源 #语音AI #Microsoft #TTS #ASR
Nav Toor@heynavtoor

🚨 Microsoft just open sourced a voice AI that was too dangerous to keep live. They took it down. Added watermarks and safety controls. Then re-released it. For free. It's called VibeVoice. Microsoft's frontier open source voice AI. Clone any voice from 10 seconds of audio. Generate 90 minutes of multi-speaker conversation. Real-time streaming. All running locally on your machine. No ElevenLabs. No $99/month subscription. No per-minute pricing. Here's what this thing does: → Text-to-speech that sounds indistinguishable from a real human → Generate up to 90 minutes of audio in a single pass → 4 distinct speakers in one conversation with natural turn-taking → Clone any voice from just 10 seconds of audio → Real-time streaming TTS. First audio in ~200 milliseconds. → Speech-to-text that processes 60 minutes of audio in one pass → Identifies who said what and when. Speaker labels + timestamps. → Supports 50+ languages for transcription → Custom hotwords for names, technical terms, domain-specific accuracy Here's the wildest part: Give it a podcast script. It generates a full multi-speaker conversation that sounds like two real humans talking. Natural pauses. Emotional nuance. Turn-taking. 90 minutes. One command. Microsoft had to take this repo down once because people were misusing it for deepfakes and disinformation. They brought it back with embedded watermarks, audio disclaimers, and safety controls. That's how powerful this is. A $3 trillion company built it. Released it. Pulled it. Fixed it. And gave it back to the world. ElevenLabs: $99/month. Play.ht: $39/month. Amazon Polly: pay per character. This: Free. Local. MIT License. 23.5K GitHub stars. 2.6K forks. Backed by Microsoft Research. 100% Open Source.

中文
27
594
2.7K
326.4K
ego6 รีทวีตแล้ว
Erick
Erick@ErickSky·
🚨 UN GENIO ACABA DE HACER HISTORIA 🔥 El crack @dx8152 (大雄) se hartó de que la gente normal no pudiera usar LTX 2.3… Y creó LTX-2 STUDIO: ✅ Versión desktop 100% LOCAL y GRATIS ✅ Rompe el límite de 32 GB VRAM (¡corre perfecto en tu 4090!) ✅ Todo en una interfaz limpia y pro: • Texto → Video • Imagen → Video • Lip-sync perfecto • Mejora de video • Frames de inicio y fin Lo que hizo este muchacho es democratizar la IA de video. Ya no hace falta pagar suscripciones caras ni tener un supercomputador. Ahora cualquiera con una GPU decente puede generar videos brutales desde su casa. Esto es lo que pasa cuando alguien piensa en la comunidad ❤️ LINKS AQUÍ👇
Erick tweet media
Español
37
267
1.9K
85.9K
ego6 รีทวีตแล้ว
John
John@johnAGI168·
看着都疼🤣 扇耳光电影片段🎬 Seedance 2.0 文生视频 prompt 👇 生成一段15秒极具戏剧张力的电影级写实剧情短片,展现一个女人在遭遇连续暴力时的绝不屈服,注意情绪递进。 镜头语言与视觉风格:全程保持绝对的固定机位近景特写(LOCKED-ON SHOT),无推拉摇移,将视觉焦点强制锁定在红衣女子的面部。背景为大面积冷灰色调的暗黑地下车库,与女子身上的高饱和度红色蕾丝裙形成极为强烈的视觉冲突,整体色调压抑、紧绷。 动作与叙事节奏: 0-7秒:画面右侧(穿紫色蕾丝袖的手)先后两次扇女子巴掌。每次女子的头因受力被打偏后,都以极其缓慢的节奏转回头,用麻木、隐忍且倔强的眼神冷冷直视施暴者。 7-12秒:暴力节奏骤然加快,施暴者连续高频扇女子约五次耳光,女子头部左右摇摆,发丝散乱,动静对比达到情绪高潮。 12-15秒:动作停止,画面重新归于平静的凝视。女子脸颊微红,碎发贴脸,再次以极度缓慢的速度转头死死盯住对方,眼神透出极度的冰冷与坚不可摧的挑衅。 特殊控制指令:人物面部特征与服装保持绝对一致无跳变,受击时的物理反馈真实自然,微表情清晰细腻。配合画面节奏生成清脆的掌掴音效与极度压抑的底层氛围音,无对白。
中文
6
9
120
7.5K
ego6
ego6@kidemc·
@YaReYaRu30Life Prend 3 compte tu peut réaliser un max de vidéo et les crédit non utiliser tu peut les reporter sur un compte précis
Français
0
0
0
45
ヤレヤル
ヤレヤル@YaReYaRu30Life·
CapCutでのSeedance2.0におけるコスト単価 コスト単価を計算しよーとは思ったのですが 他のサービスと異なるところが Seedance2.0 Standard/fast共に 5S.10S.15S指定です。 5S単価 Standard120:クレ fast:100クレ 1クレ=1円なので純粋に 5秒で100円〜120円で 他と比べて安くもなく高くもない単価ですが、 一点注意が CapCutはそもそも 動画編集ソフト的なところもあり Proプラン契約でもらえるクレジットはそこまで多くありません 月2200円で800クレジット付与ですので Seedance2.0目的でやる場合 使えば使うほど単価は1クレ1円に収束しますが 編集などで使用するなど本来のCapCutの利用目的がない場合は 普通に割高ですのでご注意を🙋
ヤレヤル tweet media
ヤレヤル@YaReYaRu30Life

【CapCut×Seedance2.0】 CapCutでSeedance2.0が 使えるとのことで Baseballでのダブルプレーに挑戦 (他で触ったのもあってネタ切れ) 今回は画面の使用無し テキストから動画生成 動きはさすが、ただ細かな位置関係などがあまりうまくいかなかったところもあるので この辺りはプロンプトでどう調節するか🤔 今後も定期的に挑戦していきたいです! 最終目標は 投手の投球から 打者の打撃後の 4-6-3のダブルプレーですが 6(ショート)とランナーをうまく利用してかっこいいジャンプスローをできればなーと🙋 @capcutapp_jp #capcutCCP #capcutjapandiscord

日本語
7
20
94
24.9K
ego6 รีทวีตแล้ว
MELTEN
MELTEN@meltenx·
カルトの花嫁
MELTEN tweet mediaMELTEN tweet mediaMELTEN tweet mediaMELTEN tweet media
MELTEN@meltenx

日本語
1
220
2.9K
43.6K
Emily
Emily@IamEmily2050·
Seedance V2 I am your sister 🤣 core_style: - 80s-90s Shaw Brothers movie style - early Hong Kong wuxia drama aesthetics - nostalgic Chinese wuxia film - vintage TV quality - warm and high-saturation color palette - retro film grain texture - slight Technicolor chromatic aberration - classic studio backdrop feel - soft stage lighting visual_quality: - 35mm film photography - physical film defects - vintage film texture - subtle chromatic aberration (dispersion) - soft focus effect - slight light flickering - strong bloom on highlight surfaces character_modeling: female_character: - classic 80s period drama makeup - dark eyeliner - peach-pink lip balm - exquisite braids with pink silk ribbons and flower accessories - traditional light blue and white Hanfu with floral embroidery and silk texture male_character: - classic wuxia young scholar appearance - long hair tied with a white silk ribbon reaching the waist - signature sideburns - clean-shaven face - pure white scholar/swordsman long robe (Zhiduo) script_and_storyboard: logic: sitcom editing logic with early physical transition effects scenes: - shot_1: "Close-up: A young woman looks affectionately at a black and white cow cat on a wooden table. Dialogue: 'If only you could turn into a handsome guy.'" - shot_2: "Medium shot: The cow cat is licking its paws, followed by a burst of retro white physical smoke effects in the center of the screen." - shot_3: "Transformation: The smoke dissipates, and a handsome man in a white robe appears where the cat was." - shot_4: "Over-the-shoulder shot: The woman looks pleasantly surprised and reaches out to touch the man's shoulder or chest. Dialogue: 'Wow, you really became a handsome guy!'" - shot_5: "Medium shot: The man has a serious expression, gesturing with an orchid finger while speaking enchantingly. Dialogue: 'Have you forgotten you already had me neutered? I've changed now, I'm your sister!'" - shot_6: "Close-up: The woman freezes with a shocked expression. Dialogue: 'Ah!?'" audio_and_post_production: - "Poof" transformation sound effect - nostalgic Hong Kong background music - AI voiceover with "Old Film" or "TVB dubbing style" accent
English
39
68
512
25.1K
Dave-X
Dave-X@DaveX_GraphX·
Le prochain chapitre de DBK est bientôt bouclé mais arrivera avec qq jours de retard. Certaines parties ont tout simplement été un peu plus complexe à réaliser que prévu. Bref, je fais tout mon possible pour que celui-ci soit mis en ligne mi-Avril grand maximum!
Dave-X tweet media
Français
2
2
16
453
ChienSurpris
ChienSurpris@ChienSurpris·
📢🚨 "On a monté notre boîte de nuit. On a fait faillite." En une phrase, le résumé de l'entrepreneuriat en France. Ce mec a TOUT tenté. Il a pris des risques, investi son temps, son argent, sa santé mentale. Et le système l'a broyé. En France, créer une entreprise c'est se battre contre l'URSSAF, les charges sociales à 45%, le Code du travail de 3 500 pages, les normes, les licences, les inspections. Tu montes une boîte de nuit ? Permis d'exploitation, licence IV, normes incendie, décibels, voisinage, et si ça marche pas c'est TOI qui rembourses tout. L'État ne prend aucun risque mais veut sa part dès le premier euro. 8 minutes de témoignage qui valent tous les manuels d'économie : en France, entreprendre c'est survivre MALGRÉ l'État, pas grâce à lui.
Français
32
152
639
45.1K
ego6
ego6@kidemc·
@VoxcatAI 这在不同模型之间会有所不同吗?
中文
1
0
0
239
VoxCat
VoxCat@VoxcatAI·
AI 绘画小技巧: 如果你写提示词不想捏脸调面部参数, 直接带上「中国顶级网红」、「韩国顶级网红/女团」或者「日本顶级女优」。 本质上是利用模型对这些高光标签的审美聚类,出来的样貌基本上不会丑。 亲测好用,建议直接复制去试。
VoxCat tweet media
VoxCat@VoxcatAI

刚把申鹤那组 Cosplay 海报的流向从 Flux 换到 Nano Banana 2 (香蕉模型) 和 gpt模型 跑了一遍,发现视觉逻辑完全变了。 很多人觉得GPT 只能画点插画,但昨晚实测发现,在处理这种“带有极强叙事感”的复杂 Prompt 时,它的语义理解确实是独一档: · 动作精准度:Prompt 里要求的“道诀感半展开手势”,GPT 几乎一次就给对,不穿模。 · 构图服从性:仰拍大长腿的 68% 比例控制,香蕉模型配合特定参数出的透视张力非常恐怖。 · 质感差异:Midjourney 追求的是“摄影美学”,而香蕉模型现在更倾向于一种“超现实的干净感”,画面几乎没有杂质。 有个踩坑点:用香蕉模型生这种写实人像,千万别加太重的滤镜词,否则会显得“AI 味”过浓。保持 Prompt 里的材质描述(丝缎、皮革、金属),让它在自然光影下自洽,出来的质感反而高级。 大家最近有用 Nano Banana 2 出过这种高拟真角色吗?感觉它在处理二次元真人化时的“骨骼逻辑”,是不是比以前稳多了?

中文
6
8
66
7.2K
ego6 รีทวีตแล้ว
VoxCat
VoxCat@VoxcatAI·
人像_角色设定卡_横版3x2_左2大右6头部表情 用途:上传参考图后,生成“角色设定卡/Turnaround Sheet”。 版式:左侧两张大图(全身正视+全身侧视),右侧 6 张小图(正面/背面/左右45°/开心/生气)。 约束:禁止任何可读文字/标签;统一棚拍白底与曝光;签名为右侧竖排 voxcat。 提示词(中文版) 3:2 横版角色设定卡/转面板(turnaround sheet / model sheet),纯白干净棚拍背景。以参考图人物为唯一身份锚点:脸型轮廓(下颌线、颧骨、下巴形状)、眼型、眉形、鼻梁与鼻翼、嘴唇厚薄与嘴角形状、年龄气质必须严格一致;发际线与发型尽量一致。只允许同一个角色,禁止换脸、禁止五官漂移。 版式(单张合成图,干净网格,统一光影与色彩): 左侧(约60%宽度):两张大图上下排列: 1)全身正视站姿(中性站姿,手臂自然下垂) 2)全身90°侧视站姿(中性站姿) 右侧(约40%宽度):2×3 网格六张头部小图: 1)头部正面(neutral) 2)头部背面(back of head,用于发型与头型一致性) 3)头部左45°(neutral) 4)头部右45°(neutral) 5)表情特写:开心/愉悦(happy,笑但克制不夸张) 6)表情特写:生气/愤怒(angry,眉眼紧张但不夸张变形) 质感与画质:高端写实棚拍/电影级人像质感,眼睛清晰锐利对焦,真实皮肤微观质感(毛孔与细纹,不磨皮不塑料),全图各分区曝光与色彩一致,8K细节,轻胶片颗粒,超干净白底,脚下干净柔和投影。 强约束:画面内不允许任何可读文字(不要 FRONT/SIDE 等标签),不要字幕、不要logo、不要UI叠层、不要水印块;不要卡通二次元;不要多余人物;不要畸形手指/多肢体/脸崩;六张小图必须是同一张脸同一发际线。 签名:右侧边缘添加极小竖排签名“voxcat”,超细字重,像艺术家署名(不是水印),不抢主体。
VoxCat tweet mediaVoxCat tweet mediaVoxCat tweet mediaVoxCat tweet media
VoxCat@VoxcatAI

x.com/i/article/2032…

中文
26
166
899
70.7K
underwood
underwood@underwoodxie96·
I didn't expect so many people to like this prompt. I've organized it into the AI ​​Effect tab, so you don't need to remember the prompts. Just go to our website and click this button to use it directly. promptsref.com/tool/AI-Image-…
underwood tweet mediaunderwood tweet mediaunderwood tweet media
underwood@underwoodxie96

This prompt can help you generate three-view character sheets, allowing AI to create more detailed and refined characters from those views. 1. Upload character image 2. Use nano banana pro and the following prompts on our website. promptsref.com/library/nano-b… A high-definition, clean, minimalist character design board / character turnaround reference sheet, set against a pure white background. The overall presentation should resemble a professional game art character modeling sheet, fashion design reference page, character design sheet, or character turnaround board. The layout should be neat and well-organized, with clearly divided information sections, a realistic and premium visual quality, consistent lighting, and strict character consistency throughout. On the left side of the composition, show the character’s full-body three-view turnaround, occupying the main visual area, including: 1. Front full-body standing pose 2. Left-side full-body standing pose 3. Back full-body standing pose All three figures must be the exact same character, with identical facial features, hairstyle, clothing, body shape, and height proportions. The standing pose should feel natural, with both arms hanging naturally at the sides. This should be suitable as a character modeling reference. The camera angle should be eye level, with neutral studio lighting, no obstruction, no exaggerated perspective, and no complex background. The right side of the composition should be divided into two sections: In the upper-right section, place six headshot / head-angle reference images of the same character, arranged neatly to show different head perspectives, including: - Front-facing portrait - Slight downward angle showing the top of the head - Back of the head / rear head view - Left-side facial profile - A near-side-angle comparison view (this line was slightly unclear in the source and may not be exact) - 3/4 profile portrait The head references should have clear facial features, visible hair parting, and consistent facial structure, making them suitable as head design references. In the lower-right section, place six close-up detail images of the character, arranged into a clean grid, showing key design details, including: - Close-up of the upper garment fabric texture - Front close-up of the lower-body clothing - Close-up of the hip / tailoring detail - Close-up of the leg or skin texture detail - Close-up of the eyes or facial feature details - Full close-up of the shoes as a standalone item All detail images must match the main character’s outfit and appearance exactly. Materials should look realistic, and the details should be clean and precise, suitable as clothing and accessory modeling references. Overall style requirements: Minimalist, professional, realistic, unified, clean, and premium, similar to a character design board, fashion design reference sheet, 3D character modeling reference page, or character turnaround presentation board. The character edges should be sharp, garment shapes should be clearly defined, hair strands should appear natural, skin should look refined, and material rendering should be accurate. The overall layout should have generous white space, as if it were made by a professional concept art team. Character setup: [Character appearance description] Output requirements: Landscape composition, white background, full character visible, no cropping, no extra props, no explanatory text, no logo, no watermark, no UI interface elements, no like/save buttons, and no social-media-screenshot appearance.

English
9
112
949
88.3K
ego6
ego6@kidemc·
@iamsupersocks C’est presque autant voir pire pour la génération de vidéo
Français
1
0
1
57
ego6 รีทวีตแล้ว
Travis Davids
Travis Davids@MrDavids1·
I've tested @dreamina_ai Seedance 2.0 to see if it can generate a commercial/spec ad from 2x2 grids or individual frames using Omni. The Omni feature is very powerful! From testing, it's certainly possible to generate a 15 second ad. I wanted to test something fashion related and another commercial for a perfume. All audio is straight from Seedance 2.0. #DreaminaSeedance2 #DreaminaAI #DreaminaCPP #Seedance2
English
16
68
543
26.9K
ego6 รีทวีตแล้ว
معلومة |Information
ابداع أبداع خيالي 🧠😱 من الزجاجة إلى الجمال قوة إعادة تدوير الزجاج.🔧
العربية
41
338
4.7K
2.4M
ego6 รีทวีตแล้ว
Theoretically Media
Theoretically Media@TheoMediaAI·
Dreamina Seedance 2.0 really CAN do it all. I had early access via the @Dreamina_ai CPP and couldn't resist pushing it as hard as I could! Presenting: Dragon Blue, an action-packed tale of vengeance and katanas! Really proud of this one and honored to showcase  the model’s capabilities on Dreamina AI. Dreamina Seedance 2.0 is now available on both Dreamina AI WEB and APP. As you likely know, it's currently rolling out in select countries and regions only. That'll change soon, so keep an eye out for official announcements! #DreaminaSeedance2 #DreaminaAI #DreaminaCPP #Seedance2 #Seedream5Lite
English
56
31
349
19K
ego6 รีทวีตแล้ว
underwood
underwood@underwoodxie96·
Some friends told me that 4K images can help AI video models produce better results than 2K images. So we’ve updated the feature: Nano Banana Pro and Nano Banana 2 now support 4K image generation. You can now choose the image resolution that best fits your needs. promptsref.com/tool/AI-Image-…
underwood tweet mediaunderwood tweet mediaunderwood tweet media
underwood@underwoodxie96

This prompt can help you generate three-view character sheets, allowing AI to create more detailed and refined characters from those views. 1. Upload character image 2. Use nano banana pro and the following prompts on our website. promptsref.com/library/nano-b… A high-definition, clean, minimalist character design board / character turnaround reference sheet, set against a pure white background. The overall presentation should resemble a professional game art character modeling sheet, fashion design reference page, character design sheet, or character turnaround board. The layout should be neat and well-organized, with clearly divided information sections, a realistic and premium visual quality, consistent lighting, and strict character consistency throughout. On the left side of the composition, show the character’s full-body three-view turnaround, occupying the main visual area, including: 1. Front full-body standing pose 2. Left-side full-body standing pose 3. Back full-body standing pose All three figures must be the exact same character, with identical facial features, hairstyle, clothing, body shape, and height proportions. The standing pose should feel natural, with both arms hanging naturally at the sides. This should be suitable as a character modeling reference. The camera angle should be eye level, with neutral studio lighting, no obstruction, no exaggerated perspective, and no complex background. The right side of the composition should be divided into two sections: In the upper-right section, place six headshot / head-angle reference images of the same character, arranged neatly to show different head perspectives, including: - Front-facing portrait - Slight downward angle showing the top of the head - Back of the head / rear head view - Left-side facial profile - A near-side-angle comparison view (this line was slightly unclear in the source and may not be exact) - 3/4 profile portrait The head references should have clear facial features, visible hair parting, and consistent facial structure, making them suitable as head design references. In the lower-right section, place six close-up detail images of the character, arranged into a clean grid, showing key design details, including: - Close-up of the upper garment fabric texture - Front close-up of the lower-body clothing - Close-up of the hip / tailoring detail - Close-up of the leg or skin texture detail - Close-up of the eyes or facial feature details - Full close-up of the shoes as a standalone item All detail images must match the main character’s outfit and appearance exactly. Materials should look realistic, and the details should be clean and precise, suitable as clothing and accessory modeling references. Overall style requirements: Minimalist, professional, realistic, unified, clean, and premium, similar to a character design board, fashion design reference sheet, 3D character modeling reference page, or character turnaround presentation board. The character edges should be sharp, garment shapes should be clearly defined, hair strands should appear natural, skin should look refined, and material rendering should be accurate. The overall layout should have generous white space, as if it were made by a professional concept art team. Character setup: [Character appearance description] Output requirements: Landscape composition, white background, full character visible, no cropping, no extra props, no explanatory text, no logo, no watermark, no UI interface elements, no like/save buttons, and no social-media-screenshot appearance.

English
0
36
318
24.9K
ego6 รีทวีตแล้ว
Guillaume Champeau
Guillaume Champeau@gchampeau·
Chose promise, chose due. Le code source est disponible : github.com/gchampeau/dysm… (si vous voulez et pouvez mettre des stars Github c'est cool, ça aidera à faire connaître hors de France. J'ai traduis l'appli en anglais et espagnol)
Guillaume Champeau@gchampeau

👉Mon fils étant dysgraphique et dyspraxique, il doit utiliser un PC en cours plutôt que d'écrire à la main. C'est compliqué pour les maths et la géométrie, et on n'a trouvé aucun logiciel facile à utiliser. Donc j'ai créé une appli web dédiée : dysmaths.com (...)

Français
23
262
645
29.6K