СПЕШУ АИ: Проводники в ИИ | ChatGPT, Claude, Gemini и ещё 300+ нейронок
3.09K subscribers
331 photos
191 videos
3 files
250 links
SpeShu.AI — платформа с мультидоступом к 300+ нейросетям

Открыть сайт 👉 https://speshu.ai
Download Telegram
Как Claude, но дешевле? Новый Grok 4.5 вышел, и вот что мы хотим сказать

Главный козырь — экономика. Grok 4.5 работает на скорости около 80 токенов в секунду, а типовую задачу из SWE-Bench Pro решает в среднем за 15 954 выходных токена против 67 020 у Opus 4.8 — примерно в 4,2 раза экономнее. В API цена $2 за миллион входных токенов и $6 за миллион выходных — заметно ниже, чем у моделей класса Opus.

Плюсы:
Кодинг от промпта до готового приложения. Модель хорошо собирает работающие приложения даже по краткому описанию. В демонстрациях xAI по одному запросу генерирует полноценную 3D-симуляцию Солнечной системы на Three.js с управлением скоростью времени и оформленным интерфейсом.
Агентные задачи. Grok 4.5 планирует и выполняет длинные многоэтапные процессы, а не просто отвечает на один вопрос — это её основное назначение в паре с агентом Grok Build.
Инженерия, наука, математика. Модель обучали на данных из программирования, науки, инженерии и математики, с упором на эффективные рассуждения под реальные рабочие задачи.


Минусы:
Старт только для разработчиков. На запуске Grok 4.5 доступна в Grok Build, в Cursor и через API-консоль, но в обычном чате на grok.com и в приложениях её пока нет. В ЕС модель обещают к середине июля.
Нет независимых бенчмарков. Все опубликованные цифры — внутренние, от самой xAI. Стороннего подтверждения «уровня Opus» на момент релиза не существует, и в этой индустрии между заявленными и независимо измеренными результатами регулярно находят расхождения. Читать стартовые цифры стоит как «данные вендора», а не как решённый вопрос.
Не лидер по сырым баллам. Даже по бенчмаркам, которые выбрала сама xAI, Grok 4.5 не занимает первого места — по большинству тестов впереди Claude Fable 5.
Репутация. ИИ-инструменты xAI пока не стали массово популярны ни у обычных пользователей, ни у корпоративных клиентов — релиз в том числе про то, чтобы это изменить.


Попробовать Grok 4.5 без дорогой зарубежной подписки за 30$ уже можно на SpeShu.AI cо скидкой 15%.

Удвойте выгоду с промокодом TG15 — он даст ещё 15% вам на счёт 🎁
👍4⚡2💯2
ChatGPT-5.6 Sol убил Claude Fable 5?!

OpenAI ударила по Anthropic, своему заклятому конкуренту. Компания впервые выпустили целое семейство моделей: ChatGPT-5.6 Sol, Terra и Luna.

Sol — флагман для самых тяжёлых задач: сложные рассуждения, длинные агентные сценарии, программирование, наука, кибербезопасность, работа с компьютером и профессиональная аналитика.

Terra — сбалансированная модель на каждый день. По заявлению OpenAI, держит уровень прошлого флагмана GPT-5.5 примерно за половину цены. Это новый вариант по умолчанию для потоковых задач.

Luna — самая быстрая и дешёвая, под высоконагруженные и чувствительные к скорости сценарии: чат-боты, классификация, черновики. При этом «самая дешёвая» не значит «самая слабая» на каждой задаче.

Вместе с семейством появилось два новых режима: max выделяет модели больше времени на размышление в рамках одной цепочки рассуждений; ultra «распараллеливает» работу между суб-агентами под сложные многошаговые задачи.

Помимо того ChatGPT-5.6 Sol обошла Claude Fable 5 в трёх тестах. OpenAI не выложили только результаты по SWE-Bench Pro. Возможно, в этом бенчмарке новый флагман всё же уступает разработке Anthropic.

Но это всё равно впечатляет, ведь ChatGPT-5.6 Sol стоит в 2 раза дешевле Claude Fable 5 на входящих токенах и на 20$ выгоднее на исходящих.

🎁 Попробуйте ChatGPT-5.6 Sol ещё экономнее на SpeShu.AI со скидкой 20%. А удвоить выгоду поможет промокод TG15.
😱3💯2🤯1
This media is not supported in your browser
VIEW IN TELEGRAM
Успейте до закрытия Claude Fable 5

19 июля Anthropic обещает скрыть эту модель из подписок. Пока не поздно, успейте повторить тренд с мультиками, навайбкодить своё приложение или прогнать большую базу данных.

Это самый мощный в мире ИИ, который действительно выполняет работу не хуже (а иногда и лучше) человека 🤯

Используйте Fable 5 со скидкой 30% на SpeShu.AI. Умножайте выгоду с промокодом TG15 и не забывайте про новые подписки.

Они включают SpeShu Claude 👉 подробности
⚡2🤩1💯1
This media is not supported in your browser
VIEW IN TELEGRAM
Раскрашиваем свой город

Этот промпт для Seedance 2.0 добавляет +100 к креативности. Вот как повторить точь-в-точь прикреплённое видео:

[Style] Live-Action + Flat 2D Anime Sticker Composite, first-person POV street-magic vlog. 8K photoreal detail with the texture of real smartphone rear-camera footage; strong contrast between the photoreal city and the flat cartoon characters. One continuous handheld phone shot. No cuts. No scene transitions.

[Camera] Raw unstabilized handheld phone footage the whole way: walking bounce, slight arm sway, occasional autofocus hunting, real exposure shifts between bright sky and building shade, slight rolling-shutter jello on fast swings. Natural phone HDR color, muted realistic tones. Between targets the camera moves in quick whip pans that follow the pen.

[Light] Late-afternoon sun from one consistent direction; every real object and every animated character drops a soft contact shadow matching that sun.

[Scene] One continuous Korean city block walked end to end: an elevated subway track overhead at the start, a tree-lined sidewalk with pigeons, a bus lane along the road, and a bus-stop bench further down. Lived-in everyday street, background passersby with natural motion.

[The Pen — magic law] The vlogger's real hand (photorealistic, visible pores) holds a black marker pen, always in frame — the visual guide connecting every beat. Every transformation follows the same sequence: the pen points, the vlogger's voice says "Biu!", blue hand-drawn sketch lines wrap the target, ink spreads, and the target becomes a lively flat-2D cel anime character with bold cartoon outlines — keeping the original object's exact size, position, speed, direction and perspective, perfectly anchored to the real street. Characters keep their flat sticker shading, never re-lit by real-world lighting; everything they touch reacts with real physics.

[00:00-00:03] Walking POV under the elevated track. The hand raises the marker into frame and points up at a passing subway train. "Biu!" — sketch lines wrap one car, and a tiny comic-style superhero swings along the train on a web line before vanishing around the corner. The train itself stays fully photoreal, same speed.
SFX: train rumble + marker whoosh + a tiny heroic swish.

[00:03-00:06] The pen swings down, camera following, to a pigeon on the pavement. "Biu!" — the pigeon becomes a cute hand-drawn cartoon bird; it hops twice, flaps, and flies off leaving little sketch feathers that dissolve into ink particles drifting toward the road.
SFX: wing flutter + papery feather rustle.

[00:06-00:10] The pen follows the particles up to the bus lane, where a city bus is passing. "Biu!" — a fast sketch outline covers the bus, and it becomes a gigantic flat-2D orange tabby cat that keeps the bus's exact size, speed, direction and perspective, calmly padding down the bus lane. Its ears twitch, its tail sways, its paws land with soft contact shadows on the asphalt.
SFX: bus engine melting into giant padded footsteps + a deep purr.

[00:10-00:12] The pen swings to the bus-stop bench where a woman sits. "Biu!" — sketch lines trace her, and the same woman, same pose, same proportions is redrawn as a vibrant anime cel illustration. She smiles warmly and waves once; a few comic sparkles drift up.
SFX: quick sketch scribble + a light sparkle chime.

[00:12-00:15] The vlogger stops and, with one final loudest "Biu—!", flings the marker spinning into the sky. The camera tilts up after the spinning pen as it draws a blue ink spiral; the ink blooms until the whole real sky becomes a hand-drawn anime sky — cel-shaded gradient, clouds with bold cartoon outlines, a hand-drawn sun. The photoreal city skyline stays fully real along the bottom of the frame. A small handwritten "THE END!" doodle — the only text on screen — pops in among the drawn clouds, and the frame freezes.
SFX: a sharp throw whoosh + a spreading watercolor wash + a felt-pen squeak writing "THE END!".
⚡3🤩3👍1💯1
СПЕШУ АИ: Проводники в ИИ | ChatGPT, Claude, Gemini и ещё 300+ нейронок
Раскрашиваем свой город Этот промпт для Seedance 2.0 добавляет +100 к креативности. Вот как повторить точь-в-точь прикреплённое видео: [Style] Live-Action + Flat 2D Anime Sticker Composite, first-person POV street-magic vlog. 8K photoreal detail with the…
Хотите раскрасить локации своего родного города?

Скорректируйте промпт в Claude. Отправьте фото своей локации и промпт из предыдущего поста. Попросите Sonnet 5 или Fable 5 сделать аналогичный промпт именно под ваше окружение.

И Claude, и Seedance доступны на SpeShu.AI без переплат за долларовые подписки. А ещё выгоднее с бонусом +15% по промокоду TG15
👍5⚡4🤩2
Media is too big
VIEW IN TELEGRAM
Как сгенерировать мультик с котиком-ниндзя? Шаг 1-й

Генерируем кота в ChatGPT Image 2.0:
Create a premium vertical 9:16 character reference sheet for one original animated character.

This image will be used as a strict identity reference for multiple consecutive cinematic scenes. The character must remain completely identical in every view.

CHARACTER:
A small, chubby black cat ninja with cute feline proportions:
- short legs;
- large round head;
- compact pear-shaped body;
- soft deep-black needle-felt fur;
- one tiny white spot directly under the chin;
- large expressive amber eyes;
- small triangular ears;
- one medium-length fluffy black tail;
- rounded natural cat paws, never human hands.

COSTUME:
The cat wears a handmade ninja outfit:
- matte charcoal-black woven fabric suit, clearly distinguishable from the black fur;
- fitted ninja hood framing the face;
- two precise openings for the ears;
- tiny dark red fabric sash around the waist;
- black fabric wraps around all four paws;
- no weapons, no shoes, no additional accessories.

VISUAL STYLE:
Premium handcrafted needle-felt stop-motion character. Visible soft wool fibers, miniature stitched costume details, realistic handmade materials, expressive puppet face, sophisticated cinematic animation quality. Cute proportions combined with the extremely serious attitude of a professional ninja. Original high-end animated short-film design.

REFERENCE SHEET LAYOUT:
Clean warm-grey studio background with soft neutral lighting.

Show exactly six consistent views of the SAME cat:
1. Large full-body front view in a neutral standing pose.
2. Full-body three-quarter view.
3. Full-body left-side profile.
4. Full-body back view clearly showing the hood, red sash, and one tail.
5. Low ninja crouching pose, preparing to jump.
6. Close-up portrait with a shocked, slightly cross-eyed comedic expression.

All six depictions must have the exact same:
- face;
- eye color;
- white chin spot;
- head-to-body ratio;
- fur texture;
- costume construction;
- red sash;
- paw wraps;
- tail length.

STRICT CONSTRAINTS:
- This is one character shown from different angles, not six different cats.
- Exactly two ears, four legs, and one tail.
- No human anatomy or human fingers.
- Do not change costume details between views.
- Do not add a room, scenery, props, laser dots, text, labels, logos, borders, or watermarks.
- Keep the character fully visible and unobstructed in every full-body view.
- Sharp, clean, high-detail production reference sheet.
🤩4👍3🤯2
Как сгенерировать мультик с котиком-ниндзя? Шаг 2-й

Затем прикрепляем кота и генерируем локацию в той же нейросети:
Use the attached cat character sheet ONLY as a reference for the handcrafted needle-felt stop-motion visual style, material quality, scale, lighting, and animation world.

Create ONE single horizontal 16:9 master location reference image.

Do not include the cat or any other character.

LOCATION:
One small rectangular traditional Japanese living room at night, constructed as a premium handcrafted miniature stop-motion set.

FIXED PHYSICAL FLOOR PLAN:

BACK WALL:
A large white-paper shoji screen centered on the back wall. It consists of exactly three equal vertical sections with a dark brown wooden grid. All three sections are closed, clean, intact, and clearly visible. Cool blue moonlight softly enters through the paper.

LEFT WALL:
One open dark-wood doorway leading into a dim corridor. This is the only doorway in the room. It must be clearly visible on the far-left side of the image.

CENTER:
One low rectangular dark-wood table standing permanently in the middle of the tatami floor. Its long side is parallel to the back wall. Leave enough open space underneath the table for the small cat to roll beneath it.

RIGHT WALL:
One low dark-wood cabinet positioned against the right wall. A round beige ceramic vase stands on the left end of the cabinet. A small green plant in a simple dark pot stands on the right end.

CEILING:
One round cream-colored hanging lamp positioned directly above the center of the low table. The lamp is switched off.

FLOOR:
Pale beige tatami mats with clean, clearly visible seams forming a logical fixed floor pattern. Leave a clear movement route from the doorway on the left, underneath the table, and toward the central shoji screen.

CAMERA:
Locked wide master camera angle.
Camera positioned exactly in the center of the front wall, approximately 50 centimeters above the tatami floor, looking straight toward the back wall.
Natural 24mm cinematic wide-angle lens.
The left doorway, central table, full three-section shoji screen, right cabinet, vase, plant, ceiling lamp, and open floor route must all be visible simultaneously.
Straight vertical architecture, no extreme perspective distortion, no Dutch angle.

LIGHTING:
Nighttime scene.
Soft cool blue moonlight comes only through the back shoji screen.
Very subtle warm ambient fill allows the furniture and tatami floor to remain visible.
Cinematic shadows, atmospheric depth, but not excessively dark.

VISUAL STYLE:
The exact same premium handcrafted needle-felt stop-motion world as the attached character. Visible miniature wood grain, handmade paper texture, soft fabric and felt details, realistic practical-set lighting, cinematic depth of field, expensive animated short-film quality.

STRICT CONSTRAINTS:
- One room and one camera view only.
- Not a collage or storyboard.
- No cat, woman, human, animal, silhouette, or reflection.
- No red laser dot or red glowing object.
- No additional doors, windows, furniture, lamps, plants, chairs, cushions, decorations, or wall art.
- Do not mirror the room.
- Doorway stays on the left.
- Cabinet stays on the right.
- Table stays in the center.
- Shoji screen stays on the back wall.
- Keep the room clean, minimal, logical, and easy to recognize from other camera angles.
- No text, labels, logos, or watermarks.
Как сгенерировать мультик с котиком-ниндзя? Шаг 3-й

Прикрепляем фото в Seedance 2.0 и вставляем промпт:
Create one coherent 15-second cinematic comedy animation, 9:16, 24 fps, using exactly 6 sequential shots connected by clean action cuts.

REFERENCE RULES:
@Image1 is a turnaround sheet of ONE single character. All six drawings show the same cat from different angles and must only define his appearance. Never show the character sheet, gray studio background, or multiple cats.
@Image2 is the exact environment and continuity blueprint. Reproduce this same Japanese room throughout the entire video: doorway on the left, three blue shoji panels on the back wall, low rectangular wooden table in the center, hanging round paper lamp above, wooden cabinet on the right with beige vase and green plant, unchanged tatami pattern and furniture placement.

CHARACTER LOCK:
One short, chubby black ninja cat identical to @Image1: round body, black fur, huge amber eyes, tiny white chin patch, black hood and ninja outfit, dark-red tied belt, short legs, one fluffy tail. Preserve the exact face, proportions, costume, colors and texture in every shot. Premium tactile handcrafted stop-motion-inspired 3D animation with physically believable movement.

TIMELINE:

0.0–2.0 seconds — Close-up from inside the room toward the left doorway. The ninja cat slowly peeks around the wooden doorframe, gripping it with both paws. He narrows his eyes suspiciously and carefully scans the room. Very slow cinematic dolly-in.

2.0–4.5 seconds — Wide shot matching the exact room layout in @Image2. The cat enters through the left doorway and tiptoes toward the central table with exaggerated silent ninja steps, knees raised, arms poised, tail balancing behind him. The room remains completely still. Subtle slow camera push-in.

4.5–7.0 seconds — Medium low shot. One tiny sharp red laser dot appears flat on the tatami directly in front of the cat, beside the front-left area of the table. The cat instantly freezes. His eyes lock precisely onto the dot, pupils enlarge, ears rise, head follows it. The dot smoothly slides across the tatami toward the space beneath the table. No visible laser beam.

7.0–9.5 seconds — Floor-level tracking shot. The cat drops into an intense ninja crawl and scurries beneath the same table, trying to catch the red dot with alternating paws. The dot always remains flat on the tatami, only a few centimeters beyond his paws, moving naturally toward screen-right. The camera trucks right at floor level while the furniture stays fixed.

9.5–12.0 seconds — Clean comedy cut. The paper lamp switches on, changing the room from cool blue night lighting to warm amber light. An adult woman kneels in the left foreground with her back completely toward the camera; her face is never visible. She holds a small black laser pointer aimed at the cat. The cat sits upright in the center in front of the table. One tiny red dot is positioned exactly in the center of his forehead. He becomes cross-eyed while staring upward at it, mouth slightly open, frozen in ridiculous confusion.

12.0–15.0 seconds — The woman quickly moves the laser dot from the cat’s forehead onto the exact center of the middle shoji panel. Several tiny red dots briefly dance across that same panel. The cat snaps his gaze toward them, launches forward in one powerful comic leap, clears the table and crashes headfirst through the center shoji screen. Paper tears outward physically. End with the cat stuck halfway through the torn screen: only his round backside, hind paws and one tail visible inside the room. Hold the final funny pose for 0.6 seconds.

CAMERA AND CONTINUITY:
Use intentional cinematic cuts, not morphing or dissolves. Maintain correct spatial geography and screen direction. Every action must lead logically into the next shot. Keep the table, doorway, shoji panels, lamp, cabinet, vase, plant and tatami in their exact original positions and sizes.

AUDIO:
No dialogue and no narration. Quiet playful Japanese stealth music, soft padded footsteps, tiny laser-pointer clicks, cloth rustle, quick whoosh during the leap, loud paper rip and wooden thump on impact, then a short comedic silence.
Media is too big
VIEW IN TELEGRAM
Как сгенерировать мультик с котиком-ниндзя? Шаг 4-й

Продолжение промпта:
STRICTLY AVOID:
multiple cats, duplicated limbs or tails, costume changes, character redesign, changing room geometry, moving furniture, extra doors or windows, changing props, woman turning toward camera, visible woman’s face, floating red orb, glowing projectile, visible laser beam, dot appearing in midair, teleportation, morphing between shots, slideshow motion, random camera shake, text, captions, logos or watermarks.


Обе нейросети доступны на SpeShu.AI без дорогих зарубежных подписок за 55$. Ещё выгоднее с промокодом TG15 — получите +15% к пополнению 🎁
⚡3👍2🤯2
Media is too big
VIEW IN TELEGRAM
Реклама для бренда батончиков на 1 МЛН просмотров

ChatGPT Image 2.0 + Seedance 2.0 творят чудеса. Правда, чтобы с одного промпта получить топовое видео, нужно создать множество кадров

Нам понадобилось целых 9 промптов: https://telegra.ph/Kak-povtorit-virusnuyu-reklamu-batonchika-v-nejrosetyah-Prompty-07-16

А чтобы оживить все эти фото, понадобился только один большой запрос. Первая часть:

Create a 15-second vertical 9:16, 24 fps, hyper-realistic premium chocolate commercial.

Use @Image1 as the exact opening state. Use @Image2, @Image3, @Image4, @Image5, @Image6, @Image7 and @Image8 as chronological visual keyframes in this exact order. Each image represents a new physical stage of one continuous product transformation.

Do not create a slideshow. Do not simply crossfade between the reference images. Turn every still image into real physical motion using motivated material transitions, dynamic camera movement, speed ramps, foreground wipes and precise match cuts.

Recompose the mixed-aspect-ratio references into a consistent vertical 9:16 composition by naturally extending the existing dark charcoal background above and below. Keep every chocolate bar and package completely visible. Never crop, stretch or distort the product.

0.0–1.5 seconds — OPENING CRACK:
Begin exactly from @Image1. Extreme macro camera slowly tracks forward along the glowing caramel fissure in the dark chocolate surface. The chocolate subtly vibrates under internal pressure. The crack gradually widens, tiny crumbs lift from the surface, warm caramel bubbles and pulses inside. Realistic shallow depth of field and microscopic chocolate texture.

1.5–3.1 seconds — CARAMEL ERUPTION:
The internal pressure suddenly releases. The camera performs a fast controlled dolly backward as the caramel erupts upward, naturally transforming into @Image2. The thick caramel column twists with realistic weight and elasticity. Chocolate fragments rotate outward in slow motion while small droplets catch the copper backlight. Add a subtle clockwise camera orbit, not a static shot.

3.1–4.9 seconds — CRUNCHY CORE ASSEMBLY:
The rising caramel spiral continues without stopping. As the camera follows it upward, crispy wafer layers and crunchy cereal pieces converge in the center and assemble into the exact floating core shown in @Image3. The caramel ribbon coils around the core in a controlled spiral. Crumbs orbit with believable inertia. Preserve the exact rectangular proportions and internal layer structure.

4.9–6.9 seconds — CHOCOLATE COATING:
Two glossy dark-chocolate waves rush in from opposite sides. They pass close to the lens, creating a natural liquid occlusion transition into @Image4. The chocolate wraps around the floating core, fills the outer shape and settles into a clean shell. The liquid surface tightens and smooths, transforming into the exact completed unwrapped bar from @Image5. The finished bar must preserve its angular segments, faceted surface and central caramel fracture. The camera moves into a low three-quarter hero angle.

6.9–9.2 seconds — PACKAGE REVEAL:
The completed bar from @Image5 makes a slow controlled 30-degree hero rotation. Caramel, chocolate and crunchy fragments form a circular vortex around it. A matte-black wrapper sweeps in from behind and folds tightly around the bar, creating a fast but believable match transformation into the exact KRAKK package shown in @Image6. The package settles nearly front-facing while the ingredient vortex continues rotating around it.

Preserve the package perfectly. The brand name must remain exactly “KRAKK” and the smaller text must remain exactly “CARAMEL CRUNCH”. Never replace, rewrite, deform or animate the letters. Keep the logo clear and readable.
👍4🤯3⚡2💯1
Реклама для бренда батончиков на 1 МЛН просмотров. Вторая часть промпта для анимации:

9.2–11.8 seconds — MACRO SNAP:
A large foreground chocolate fragment rapidly crosses the lens and creates an impact cut into @Image7. In extreme slow-motion macro, the exact chocolate bar snaps once into two pieces. The pieces move apart while one thick glossy caramel strand stretches naturally between them. Wafer layers, cereal spheres and chocolate flakes remain sharp and recognizable. The camera makes a short lateral tracking movement with a precise rack focus from the upper half to the lower half. One powerful crisp crunch, no repeated breaking.

11.8–15.0 seconds — FINAL PACKSHOT:
The caramel strand sweeps across the lens as a warm amber wipe. The camera smoothly pulls back and resolves into the exact final composition of @Image8.

The intact KRAKK package stands behind the bar while the broken piece settles gently in the foreground. A few crumbs fall onto the surface and the remaining particles slowly lose momentum.

Preserve the exact package design, product proportions, cross-section and lighting from @Image8. Hold the final packshot almost still for the last 1.2 seconds, with only subtle drifting particles, a soft copper glow and a very slow final camera push-in.

VISUAL STYLE:
Ultra-photorealistic premium food advertising, dark charcoal-black and molten copper palette, realistic caramel viscosity, believable chocolate physics, crisp wafer texture, shallow depth of field, controlled motion blur, dramatic side and rim lighting, polished high-end CGI combined with macro commercial photography.

SOUND DESIGN:
Dark premium electronic rhythm with deep restrained bass pulses synchronized to each transformation. Detailed chocolate cracking, sticky caramel stretching, flying crumbs, liquid chocolate movement, one powerful wafer crunch and a subtle wrapper snap. Finish with a deep clean product-impact sound. No voiceover and no dialogue.

STRICTLY AVOID:
No slideshow, no static Ken Burns effect, no slow dissolves, no random objects, no duplicate bars, no duplicate packages, no product deformation, no melting geometry, no incorrect internal layers, no altered logo, no misspelled text, no extra text, no subtitles, no watermark, no hands and no people.


Используйте ChatGPT Image 2.0 и Seedance 2.0 на SpeShu.AI с промокодом TG15 — получите бонусом 15% к вашему пополнению.
👍3💯3🤩2
Половина лета прошла, а у вас до сих пор нет фото с курорта?

Исправьте это в Фотостудии SpeShu.AI — готовые летние образы ждут вас. Просто отправьте своё фото, и сервис выдаст серию из трёх снимков в одной стилистике 🔥

Дополнительно вы можете примерить новую одежду, не ходя в магазин, и поэкспериментировать со своим образом. Лето — время обновления. Не упустите эту возможность — удивить родных и друзей короткой причёской или пирсингом 😁
🤩4👍2🤯2
⚡️ Какие обучающие посты написать для вас? Задайте вопросы в комментариях
👍3🤔2😱2
This media is not supported in your browser
VIEW IN TELEGRAM
Kimi K3 собрал игру про капибар за три промпта

Новый китайский флагман догоняет Claude Fable 5 и ChatGPT-5.6 Sol по производительности.

В бенчмарках Kimi K3 легко справляется с многоэтапными задачами, управляет агентами и пишет неубиваемый код. Контекстное окно — аж 3 МЛН токенов. Нейросеть часами «помнит» всё, о чём вы её попросили 🤯

При этом Kimi K3 в несколько раз дешевле американских лидеров рынка. Так что если хотите сэкономить на крупной задаче, это умный способ.

Сэкономьте ещё больше со SpeShu.AI — Kimi K3 уже на платформе. Получите бонус по промокоду TG15 и забудьте про дорогие подписки за $ и €.
⚡5👨‍💻4✍3