Annotate · Train · Publish · Work 10x faster
An AI-powered image and video annotation tool. Optimized for model training annotation, it batch-generates high-quality training datasets and quickly reverse-engineers prompts from images and videos.
An end-to-end AI model trainer with new Boogu Image, Boogu Image Edit, and KREA 2 RAW training support. Tutu timestep is built in, no parameter setup required, and LoRA training starts in one click.
Publish to major social platforms in one click. AI automatically completes titles, tags, and descriptions on a pay-as-you-go basis; all other publishing features are completely free.
A professional workstation for AI image generation, AI video, asset breakdown, model training, and prompt reuse.
Batch reverse-engineer asset prompts in phrase mode and batch-edit the generated results.
Works with 30+ major global video platforms. Reverse-engineer online videos without downloading them, then generate scene descriptions, summaries, and voiceover scripts for creative reference, asset breakdown, and model training.
Reverse-engineer image prompts as concise phrases or natural language. Process thousands of images in one run with stable, accurate long-task output.
Input up to 10 source images and a result image to describe what changed with precision, useful for editing-model training and specialized scenarios.
Generate region annotations with AI or draw boxes manually, then export Region JSON datasets for Ideogram 4 / TutuTrainer with image-level descriptions, elements, and palettes.
Mark elements in videos you have the right to edit with clicks, boxes, or brush masks. The app generates masks, tracks them across frames, and removes distracting subtitles, watermarks, logos, or other elements.
Default mode requires no local model download or API debugging. Install the app and start reverse-engineering right away with minimal system overhead.
Advanced mode supports multiple configuration methods and provider formats, so you can connect the multimodal models your task needs.
Specialized filtering adapts prompt output for training datasets, producing captions better suited to LoRA training and stronger results.
Edit, split, reorder, and organize generated prompts at scale. The neural module keeps thousands of image captions manageable.
Resize and convert images, rename assets, detect similar images, and combine preprocessing steps freely before training.
Use leading Image2 generation models with hundreds of built-in art styles for fast high-quality creation and style exploration.
A standard Windows installer: click through setup and start working, with no complicated environment configuration.
Zero-barrier onboarding, full model support, ready out of the box, rich resources, enterprise-grade stability, and completely free with no tricks or ads.
Supports common training paths including SD1.5, SDXL, FLUX, FLUX2 Klein, Z-Image, Qwen-Image, Boogu Image, KREA 2 RAW, LTX, and WAN, bringing scattered model-training workflows into one visual workstation.
No manual Python setup, dependency hunting, or launch-parameter guessing. Use the Windows installer, click through setup, then choose a dataset and model to start your first LoRA training run.
Automatically configures training parameters based on hardware, dataset, and model conditions, while optimizing VRAM usage for demanding models so you can focus on materials and results.
Uses Tutu timestep together with optimized training methods and parameters. KREA 2 RAW now has a dedicated Tutu timestep strategy, improving training results with real measurements so outcomes are more stable.
Checks model architecture, paths, and training content before launch, warning you when something looks mismatched instead of letting a bad run waste time.
View generated samples, Loss, and learning-rate curves during training, with controls for smoothing and sampling intervals so training status is easier to judge.
Prompt templates are organized by training type: characters, clothing, style, objects, and scenes, making it easier to test real-world generalization.
Connect resources from HuggingFace, ModelScope, and netdisk links, with cached model saving, single-file conversion, and custom model configuration.
Training cards show elapsed time, estimated remaining time, logs, and key status, so long-running jobs are easier to understand and monitor.
Built-in auto updates, package integrity checks, improved logs, and clearer error messages reduce uncertainty during installation, upgrades, and training.
A multi-platform video publishing tool built for self-media creators. Built-in AI autocomplete automatically generates all publishing assets and eliminates repetitive work.
Publish one video to major channels including Douyin, Kuaishou, Bilibili, WeChat Channels, TikTok, YouTube, Weibo, Baijiahao, and more from one workflow.
Add multiple accounts per platform, verify status, switch active accounts, and change account groups in one move for matrix and multi-brand operations.
Manage daily publishing slots by time, with clear pending, completed, failed, and skipped states so the content calendar is easy to scan.
Supports common formats such as mp4, mov, avi, and mkv. Queued videos can be assigned automatically to today or future empty slots for easier batch scheduling.
Generate titles, descriptions, tags, and keywords from video content, with separate language and title-style options for domestic and overseas platforms.
Run automatic pre-analysis when videos enter the queue, trigger completion manually for a single plan, or let the app fill missing copy and covers before publishing continues.
Regenerate covers for selected platforms with custom descriptions, reference images, and video understanding, then switch between historical cover versions.
Create multiple candidate titles for a platform, reuse saved tag groups, and apply templates for brand text, copyright notes, and fixed phrasing.
The dashboard shows today’s plans, AI progress, failed items, and operation logs, with retry options by platform or by scheduled slot.
Publishing features are free when AI completion is not used. AI analysis consumes credits by video length after completion, and failed analysis is not charged.
Video tutorials and product demos voluntarily published by users across major platforms
Fully open source, free to use, download from GitHub
A ComfyUI custom node for image generation that supports multiple API providers, making it easy to connect to major visual models and generate high-quality images.
Download on GitHubAn AI prompt expansion tool that supports intelligent prompt expansion for text-to-video and image-to-video, generating high-quality creative prompts in one click.
Download on GitHubA fully automated Git + GitHub hosting system, integrating version control and cloud backup into one solution so code management becomes effortless.
Download on GitHubChoose the plan that fits you and unlock all features
Publishing features are always free · AI autocomplete is billed by credits · Please recharge inside the software
Please recharge credits from the "Account & Credits" page inside the software. New registrations receive 10 trial credits.
Choose a payment method and receive an activation code after payment
$4.97 / month on average
$3.08 / month on average
After purchase, you will receive an activation code. Enter it in the software to activate.
Look up existing activation codeDetailed product tutorials and answers to common questions
AI image and video prompt reverse engineering, including activation, model access, batch processing, and export.
Install, prepare datasets, choose architectures, start training, monitor jobs, and manage checkpoints.
English docs are available. The app interface is currently Chinese-only; this guide covers scheduling, titles, tags, covers, and accounts.
Independent AI API provider checks for availability, cache hits, pricing, and public risk signals.
Questions? Contact us anytime