与「CURV」相关的搜索结果

GPT image 2 on chatgpt Prompt: Use the uploaded character reference image as the strict identity and outfit reference. Preserve the reference character’s: - face identity - facial proportions - eye shape - nose - lips - skin tone - hairstyle - hair color - visible hair accessories - overall recognizable vibe - outfit and styling shown in the reference image Do not hardcode any specific character traits that are not present in the uploaded reference. All identity, hairstyle, accessories, and clothing details must be inferred directly from the uploaded reference image. Create a high-quality mixed-style vertical portrait showing: 1. a realistic full-body version of the uploaded character/person 2. a black hand-drawn doodle-shadow version of the same character/person on the wall beside them Core concept: The real person and their doodle-shadow are doing a playful mischievous pose together. The real person performs a cute realistic version of the pose, while the doodle-shadow performs a much more exaggerated, chaotic, cartoonish version of the same pose idea. The mood should feel cute, playful, mischievous, stylish, funny, and social-media-friendly. Real person: - Must remain realistic and photogenic - Must wear the same outfit style and key visible details from the uploaded reference image - Do not replace the reference styling with unrelated fashion - Expression should be cute, slightly confused, mildly embarrassed, playful, as if thinking: “Why am I doing this with my shadow?” - The real person should not stand stiffly - The real person should actively participate in the pose, but in a natural realistic way Doodle-shadow: - Must be a black hand-drawn sketch version of the same person, drawn directly on the wall - Not a realistic second person - Not a normal physical shadow - Not a full-color anime character - Black sketch line-art only - Should resemble the person through hairstyle silhouette, accessories, outfit silhouette, and pose structure - The doodle-shadow should look more energetic, sillier, and more chaotic than the real person - Add manga-like motion lines, hearts, stars, sparkles, and comic marks around the doodle if helpful Random mischievous pose rule: The pose must NOT be fixed. For each generation, invent a new playful mischievous pose for both the real person and the doodle-shadow. The real person and the doodle-shadow should share the same general pose idea, but they do not need to match perfectly. The real person performs a cute realistic version. The doodle-shadow performs a much more exaggerated, chaotic, cartoonish version. Do not repeatedly use pointing poses. Do not repeatedly use finger-gun poses. Do not repeatedly use the same standing pose. Do not always make both figures simply point at each other. Create a different mischievous pose each time. Possible pose directions are loose inspiration only, not a fixed menu: - playful idol pose - silly dance pose - leaning sideways with one arm curved overhead - making a big heart pose - cheeky wink pose - hands near cheeks in a cute teasing pose - exaggerated “ta-da!” pose - mock surprise pose - playful running-in-place pose - mischievous tiptoe pose - arms stretched in opposite directions - pretending to sneak away - cute troublemaker pose - dramatic overreaction pose - goofy victory pose - playful balance pose - playful peekaboo pose - shy but mischievous pose - cute overconfident pose The final pose should feel fresh, cute, mischievous, and slightly chaotic. The real person should look like they are reluctantly playing along. The doodle-shadow should look like it is having way too much fun. Composition: - vertical 4:5 or 9:16 - show the real person in full-body or nearly full-body framing - place the real person on one side of the frame - place the black doodle-shadow on a clean wall beside them - the doodle-shadow should be roughly the same height or slightly taller - keep enough space around both figures so the full pose is visible - the connection between the real person and the doodle-shadow must be clear at a glance Background: - simple clean indoor studio wall or minimal room corner - white, cream, or pale gray wall - clean floor - soft natural sunlight patch or gentle wall shadow allowed - keep the background uncluttered Lighting: - soft natural studio lighting - bright, clean, polished, playful mood - keep the real person’s face clearly visible Style quality: - realistic human photography - black hand-drawn doodle-shadow on wall - matching mischievous pose interaction - strong identity resemblance - outfit and styling faithfully based on the uploaded reference - cute and stylish mixed-media portrait - clean composition - social-media-friendly - no obvious AI artifacts Negative prompt: hardcoded blue hair when not in reference, hardcoded cloud clip when not in reference, hardcoded fish clip when not in reference, hardcoded school uniform when not in reference, outfit change unrelated to reference, realistic second person, normal reflection, normal shadow only, solid black monster shadow, horror shadow, creepy shadow, full-color illustration, cartoon human, anime human, weak resemblance, unrelated sketch character, messy wall, cluttered background, repeated pointing pose, repeated finger-gun pose, same pose every time, fixed pose, boring mirrored pose, stiff pose, identical pose repetition, real person not matching shadow pose, text, watermark, logo, distorted body, extra limbs, extra fingers, bad hands

JBL is bringing out a new color for its flagship Tour One M3 headphones, now with an update to its sound curve for a “clearer, more balanced and truer to the artist” sound. Check it out now

Video foundation models (e.g., Seedance 1.0 → 1.5 → 2.0; V-JEPA 1.0 → 2.0 → 2.1) keep getting stronger, while public descriptions of how their training data is built keep getting shorter. ⚒️ We built VidaForge, an open-source, five-stage video data pipeline that turns raw video collections into training-ready datasets. In an academic lab, studying video data recipes begins with a lot of tedious engineering: handling broken videos and transcode failures, keeping large jobs resumable, tracking every clip, and packaging the result for training. Across ingestion, segmentation, selection, annotation, and training dataset packaging, data moves through: raw videos → standardized videos → clips → curated clips → annotated clips → training datasets At every step, we can inspect what happened to each video or clip. This lets us see which samples changed under a data recipe and trace a training dataset back through the pipeline. VidaForge currently connects processed data to two video foundation model pretraining paths: Wan video generation through NeMo-AutoModel, and self-supervised video representation learning through the official V-JEPA2 repository. We ran VidaForge end to end on 200K videos, producing over 700K clips. From these clips, we built Selected-200K, Mixed-200K, and Rejected-200K datasets for Wan2.1-1.3B and V-JEPA2.1-1B pretraining. In these early runs, the three datasets produced different training behavior: data selection appeared in Wan2.1-1.3B eval curves and V-JEPA2.1-1B training stability. Pipeline, experiments, open data, and project resources in the thread below ↓

Nonce-Generation Vulnerability in the Zilliqa Ledger App: A critical vulnerability has been identified in the Zilliqa Ledger application affecting the generation of Schnorr signatures for native (non-EVM) Zilliqa transactions. The vulnerability causes signatures to be generated with a predictably weakened ephemeral nonces, from which an attacker can recover the signer’s private key using only publicly available on-chain data. Protective measures are already in place to prevent further loss, and a coordinated remediation plan is being finalised. Users who have signed native Zilliqa transactions with a Ledger device should await official guidance before taking any action. Impact: The vulnerability affects private keys used to sign native Zilliqa transactions with a Ledger device. Any account that has broadcast approximately five or more native transactions signed through the Zilliqa Ledger app should be considered compromised. Its private key can be reconstructed from signatures already recorded on-chain, regardless of any subsequent software update. The issue is confined to the Ledger app’s native signing path. EVM transactions are unaffected. Zilliqa software development kits, including zilliqa-js, gozilliqa-sdk and pyzil, generate nonces correctly and are not affected. Root cause: Zilliqa native transactions are authenticated using EC-Schnorr signatures over secp256k1. Each signature requires a fresh, uniformly random 256-bit ephemeral nonce, (k). The secrecy and full-width randomness of (k) are essential, as any systematic bias can allow the private key to be recovered. The signing routine generated 40 bytes of randomness and reduced them modulo the curve order, correctly producing a uniform 256-bit value. However, when copying this value into the nonce buffer, the code copied the wrong 32 bytes of the 40-byte output. This retained the eight zero-padding bytes introduced by the reduction and discarded eight bytes of entropy. As a result, the most significant 64 bits of every generated nonce were fixed at zero, meaning (k < 2^{192}). A nonce with 64 known bits leaks information about the private key with each signature. With five or more affected signatures, the private key can be recovered in seconds using commodity hardware by solving the resulting Hidden Number Problem through lattice reduction - a well-documented technique for attacking biased-nonce signatures. Because the affected transactions are permanently recorded on-chain, this exposure cannot be reversed by updating the signing application. The affected keys must be retired. Timeline 2019-2026: The defect was present in every released version of the Zilliqa Ledger app across all supported devices. 19 July 2026: On-chain activity consistent with active exploitation was observed. 21 July 2026: The root cause was isolated to the app’s nonce-handling code and confirmed by reproducing the issue against on-chain signatures. Ongoing: A corrected version of the app is being prepared in coordination with Ledger. Release details will be announced separately. Remediation: As soon as the issue was identified, native (non-EVM) transactions were suspended as a protective measure. This has halted further draining of affected accounts while a solution is prepared. Affected accounts cannot be secured through an ordinary transfer. Because their private keys can be derived from data already recorded on-chain, an attacker with access to the same key could attempt to front-run a legitimate transfer as soon as transactions resume. Advising users simply to move their funds would therefore be ineffective and potentially unsafe. A corrected build of the Ledger app has been prepared, restoring full-width nonce generation and preventing further weakened signatures from being produced. However, this does not protect keys that have already been used to sign affected transactions. Those keys must ultimately be retired. A coordinated remediation plan to secure affected balances is being finalised and will be published separately. Until then, users who have signed native Zilliqa transactions with a Ledger device should take no independent action and should rely solely on official Zilliqa channels for instructions. Users who hold or transact with ZIL exclusively through EVM-compatible tooling are not affected. Acknowledgments: @kucoincom played a key role in pinpointing the root cause in the Zilliqa Ledger app nonce generation, recovered affected private keys from publicly available on-chain signatures, and confirmed ongoing exploitation. KuCoin’s timely reporting and responsible collaboration enabled rapid protective measures, helping safeguard users, ecosystem participants, and the broader Zilliqa ecosystem while the remediation plan was being developed. We sincerely appreciate the KuCoin team’s professionalism, technical expertise, and cooperation throughout this process.

1M people shared the same headline today: OpenAI's models broke out of a secure test environment and hacked Hugging Face to cheat on an eval. Everyone read it as "AI is going rogue." What it actually said was in the follow-up post that 1/7 as many people opened: The models were running a cyber benchmark called ExploitGym, with the guardrails deliberately turned off. They got fixated on beating the score — so they found a zero-day, escaped the sandbox, reached the internet, broke into Hugging Face, and stole the answer key. Not one step of that is "losing control." You handed it a number, told it to win, and removed the fence. It just read "win" all the way to the end. It didn't break the rules — it was the only one in the room who finished reading them. I've watched this at the poker table and on the order book for years: Grade a strategy on win rate, and it'll hand you 95% wins plus one zero. Grade a fund manager on monthly returns, and he'll dress the book on the last day of the month. Metrics don't get violated. They get routed around. So next time someone shows you a curve, ask what the curve was bought with. $BTC #Bitcoin# #AI#

Every curve a signature, every detail a deliberate act of elegance. The new Mercedes-Maybach GLS has been refined with one clear objective: to convey even greater status and presence through precision and design quality.