Loading...
Loading...
Found 111 Skills
Practical guidance for training MoE VLMs in Megatron Bridge. Compares FSDP and 3D-parallel approaches, using rounded lessons from Qwen3-VL, Qwen3-Next, and other multimodal experiments.
Integrate Modellix's unified API for AI image and video generation into applications. Use this skill whenever the user wants to generate images from text, create videos from text or images, edit images, do virtual try-on, or call any Modellix model API. Also trigger when the user mentions Modellix, model-as-a-service for media generation, or needs to work with providers like Qwen, Wan, Seedream, Seedance, Kling, Hailuo, or MiniMax through a unified API.
Return public original model architecture diagrams for user-specified LLM, VLM, MoE, diffusion, OCR, and SGLang/sgl-cookbook model families. Use when the user asks for a model structure chart, architecture diagram, or rendered image link for a specific model such as DeepSeek, GLM, Qwen, Kimi, MiniMax, Step, Hunyuan, or Qwen3-VL.
Use when the user wants to query Lore git trailers from commit history — surface constraints, rejected alternatives, directives, test gaps, and other decision context. Triggers on "최근 제약조건 확인", "리젝된 대안 보여줘", "check recent constraints", "show directives", or explicit skill invocation (/lore-query in Claude Code, $lore-query in Codex CLI, /skill:lore-query in Kimi Code, lore-query in Qwen Code).
Creates professional AI image/video prompts with photographer's and cinematographer's eye. Specializes in composition, lighting, color grading, and storytelling. Use when generating AI images/videos with artistic vision, working with models like Nano Banana Pro, Qwen, Sora2, Wan 2.2. For graphic design work (thumbnails, banners, layouts), use /graphic-designer instead.
Long-context MoE training guidance for Megatron Bridge. Covers CP sizing, selective recompute, dispatcher choices, and practical patterns from DSV3, Qwen3, and Qwen3-Next long-context experiments.
Universal AI voice / text-to-speech skill supporting OpenAI TTS (gpt-4o-mini-tts, tts-1), ElevenLabs multilingual TTS with voice cloning, Bailian Qwen TTS (qwen-tts / qwen3-tts-vd with voice-design custom voices, long-text chunking built in), MiniMax speech-02-hd, SiliconFlow CosyVoice / SenseVoice, and PlayHT 2.0. Use this skill whenever the user asks to read text aloud, synthesize speech, generate narration, create voice-over, dub a script, or turn any text into audio (mp3 / wav / ogg / flac). Typical phrases include "read this aloud", "generate voice for ...", "create a narration of ...", "tts this", "把这段念出来", "做个配音", "合成语音", or mentions of voices / TTS model names like Alloy, Ash, Cherry, Rachel, CosyVoice, PlayHT. Always use this skill even if the user does not specify a provider — pick one from EXTEND.md defaults or available env keys.
[QianWen] Check for qianwen-ai updates and notify the user when a new version is available. TRIGGER when: user asks to check for updates, check version, asks 'is there a new version', 'latest version', 'update skills', 'check update', or any other qwen skill delegates to this skill, or user explicitly invokes this skill by name (e.g. use qianwen-update-check). DO NOT TRIGGER when: non-update-related tasks, general version questions about other software.
Use when the user wants to configure Lore commit format in their project or globally — writes Lore rules to the agent's instruction file (AGENTS.md, CLAUDE.md, QWEN.md, or global agent config) so all agents automatically use structured git trailers in commit messages
Choose the right MoE token dispatcher (`alltoall`, DeepEP, or HybridEP) for the hardware, EP degree, and optimization stage. Summarizes patterns from DSV3, Qwen3, Qwen3-Next, and VLM bring-up work.
Practical guidance for training MoE VLMs in Megatron Bridge. Compares FSDP and 3D-parallel approaches, using rounded lessons from Qwen3-VL, Qwen3-Next, and other multimodal experiments.
Use Alibaba Cloud DashScope API and LingMou to generate AI video and speech. Seven capabilities — (1) LivePortrait talking-head (image + audio → video, two-step), (2) EMO talking-head, (3) AA/AnimateAnyone full-body animation (three-step), (4) T2I text-to-image (Wan 2.x, default wan2.2-t2i-flash), (5) I2V image-to-video (Wan 2.x, default wan2.7-i2v-flash, supports T2I→I2V pipeline), (6) Qwen TTS (auto model/voice by scene, default qwen3-tts-vd-realtime-2026-01-15), (7) LingMou digital-human template video with random template, public-template copy, and script confirmation. Trigger when the user needs talking-head, portrait, full-body animation, text-to-image, text-to-video, or speech synthesis.