Machine Learning

Mind Lab

Mind Lab is an industry-focused platform for model training and inference, turning enterprise expertise into proprietary model capabilities. It serves pharmaceutical, financial, public-sector, and gaming companies with end-to-end services spanning benchmarking, data development, model post-training, inference, and deployment.Your Industry Expertise, Now Trainable into Your Model.With LoRA, shared bases, and distributed optimization, Mind Lab cuts training compute cost by over 90% and starts specialist training at just $1,000. It supports frontier bases like Kimi K3, DeepSeek V4.1, Qwen3.8, GLM-5.3, and MiniMax-M3, offering three working modes: FDE experts, automated platform training, or self-service training.FeatureVertical ModelGeneral ModelSupport5×8CommunityWorkspacePrivate/VPC/HybridStandard cloudTraining LoopDedicatedBasic

210 upvotesOct 28, 2026

DeepInfer

DeepInfer is DeepSeek's open-source inference framework for serving the DeepSeek-V4.1-Flash model at scale. It unifies optimized kernels like FlashMLA, DeepGEMM, and DeepEP into a single engine with expert parallelism, speculative decoding, and sliding-window cache recovery.An engine, not a service.The release includes deterministic modes and asynchronous scheduling, but omits KV cache, prefill/decode disaggregation, and load balancing—leaving those to the user. It supports NVIDIA SM100 and Huawei Ascend 950 hardware paths.FeatureIncludedOptimized kernelsYesExpert parallelismYesKV cache systemNo

228 upvotesOct 21, 2026

manga translator

manga translator translates the text in uploaded comic images into your desired language. This AI comic translator uses OCR to recognize text and places the translation to match the original page layout. It offers batch translation of multiple images, limited free usage, and paid plans.

196 upvotesOct 28, 2026

SoundToMIDI

SoundToMIDI converts MP3, WAV, FLAC, M4A, OGG, AAC, and compatible MP4 files into editable MIDI directly in your browser. Audio transcription runs locally with no account required. Raw MIDI downloads are free, while an optional cleaned MIDI export costs $2.99 once. It works best with clear melodies or isolated instruments; dense mixes may need manual editing.

173 upvotesOct 27, 2026

AI Image Cleaner

AI Image Cleaner, AI Object Remover, Remove Person from Photo, AI Watermark Remover, Remove Text from Image, Brush Mask Eraser, AI Inpainting, Photo Cleanup, Online Object Remover, Free Guest Cleans

81 upvotesOct 26, 2026

minimax h3 Video Generator

MiniMax H3 is a multimodal AI video generation model that creates 4–15 second high-quality video clips from text, first and last frames, and image/video/audio references. Users can control motion, camera, and visual style through prompts, and choose between 768P or 2K resolution output.Key features include: text-to-video, first-frame and last-frame control, multimodal references, camera and motion guidance, and watermark-free downloads up to 2K."Create 4–15 second videos from text, frames, or references, up to 2K resolution."Ideal for product marketing, e-commerce content, social media clips, film pre-visualization, game concepts, and motion design. Offers one-time credit packs that never expire.

50 upvotesOct 26, 2026

LoRA AI train

Turn One Image Into a Consistent Wan 2.2 LoRA VideoUpload a clear start image, select a trained Wan 2.2 LoRA, describe the motion, and generate a custom 5- or 8-second video in 480p or 720p.This is an image-to-video workflow. A start image, motion prompt, and compatible trained LoRA are required. It is not a text-to-video or video-to-video editor.LoraAI 提供 Wan 2.2 图像到视频的 LoRA 生成器，用户上传起始图片、选择已训练的 LoRA 并描述运动，即可生成 5 或 8 秒的 480p/720p 视频。核心功能包括运动预览、LoRA 强度调节、负面提示词以及多 LoRA 组合（高级版）。“起始图片和 LoRA 协同工作，使主体更接近预期外观。”该工具强调视觉一致性，适用于角色、产品或艺术风格，避免长描述导致的漂移。定价基于积分，从 70 积分起，提供标准版和高级版（月度/年度）。

172 upvotesOct 25, 2026

MiniMax H3

MiniMax H3 is an advanced AI model for next-generation video creation, accessible directly in the browser via Hevzo. It supports multimodal understanding, combining text, images, video, and audio in a unified creative process. With native stereo audio and up to 15-second videos, creators can produce cinematic content with synchronized sound and smooth storytelling. The model delivers 2K video output for sharp, detailed visuals, making professional video production more accessible.Turn simple ideas into compelling videos with the MiniMax H3 video experience.Use cases include animating product photos, creating social media clips, bringing brand assets to life, and generating game trailers from concept art. Hevzo offers instant free access with daily credits and free cloud storage for all generated content.

117 upvotesOct 25, 2026

Aethra AI

Aethra AI is a pre-seed startup building a runtime defence and governance platform for AI agents. Our core product is an enforcement layer — a Planner / Quarantined Extractor / Executor architecture aligned with dual-LLM security research (CaMeL) — that tracks data provenance and blocks untrusted instructions before they can drive agent actions.Our open-source SDK (Apache 2.0, Python and Node/TS) is available on GitHub, with an optional hosted control plane for approvals, audit evidence, and kill-switch functionality.In benchmark testing on AgentDojo, Aethra blocked 35/35 injection goals with 0 false positives (full LLM-backed evaluation in progress).We're currently raising a €1.0–1.5M pre-seed round.

128 upvotesOct 22, 2026

vynoa

Vynoa is an AI-powered platform for creating videos and images from text or image prompts. It offers a unified workspace where users can generate, compare, and organize their creative outputs.Key features include text-to-video, image-to-video, text-to-image, image-to-image, and faceless video creation. The platform integrates multiple AI models like Seedance, Veo, and Kling, automatically routing each request to the best-suited model.Choose the outcome instead of wrestling with providers.Vynoa also provides focused tools for content creators, brand marketers, and memory keepers, such as AI Baby, AI Beauty Test, Deep Nostalgia, and Old Photo Restoration. All generated assets are stored in a shared library with transparent credit costs.

100 upvotesOct 8, 2026

Image Translator

Image Translator 是一款AI工具，能识别图片中的文字并将其翻译成132种以上语言，同时保持原始设计、字体和布局不变。它支持JPG、PNG和WebP格式，可一次上传最多10张图片，并自动检测源语言。翻译文字，保留设计。该工具特别适合电商产品图、菜单、漫画、截图等真实场景，提供上下文感知翻译、高质量输出和可靠处理（失败自动退款）。计划价格月额度Starter$99/年500积分Professional$199/年1500积分Business$499/年4000积分

180 upvotesSep 27, 2026

Lora AI wan video train

LoraAI offers an online service for training Wan 2.2 video LoRA models. Users simply upload a ZIP file containing at least 10 images, set a trigger word and training parameters, and complete the training entirely in their browser—no local GPU or complex environment setup required.The tool specializes in learning visual identities (such as characters, objects, or styles), and the trained LoRA can be reused for Wan 2.2 image-to-video generation. The default settings (1000 steps, learning rate 0.0002, rank 32) consume 300 credits, with support for adjusting parameters and real-time cost display."A focused dataset beats a crowded one."Once training is complete, the high-noise and low-noise weights are saved as a pair, making them ready to use directly in the generator.

160 upvotesSep 27, 2026