Max Video AI: H3 Video Maker

Version 3.0.1 introduces Reference Video, a way to create with images, videos, and audio, together with improvements to model selection, media uploads, generation settings, and task reliability. NEW: REFERENCE VIDEO • Open Reference Video directly from the home screen. • Combine images, videos, and audio in one generation task. • Use references to guide characters, subjects, movement, sound, composition, and visual style. • Mention uploaded media in your prompt and describe the role of each reference. • Receive clearer guidance when a local reference is missing, unreadable, unsupported, too short, or over a model limit. • Validate media type, quantity, duration, and availability before submitting a task. NEW REFERENCE MODELS • MiniMax H3 Reference combines images, videos, and audio, with 768p and 2K options. • Seedance 2.0 Reference supports multimedia references and quality options from 480p to 4K. • Vidu Q3 Reference uses multiple images to guide consistent characters, products, and styles. • Kling Omni O3 adds Standard, Pro, 4K, and Audio variants for different reference, quality, and sound needs. Supported ratios and 5, 10, or 15-second durations vary by model and update automatically after selection. CLEARER MODEL CAPABILITIES Upload controls, aspect ratios, resolutions, durations, capability labels, and credit estimates now respond to the selected model. Corrected several labels and request parameters to better match current service requirements. BETTER PROMPTS AND MEDIA REFERENCES • Insert uploaded media as individual prompt references. • Improved reference recognition and removal so deleted media does not leave invalid markers. • Added validation for unavailable references and improved prompt layout, scrolling, and state synchronization. IMPROVED MEDIA UPLOADS • Refined media selection, upload, preview, reordering, and removal. • Added checks for supported formats, file counts, unavailable media, cumulative duration limits, and videos that are shorter than required. • Reduced duplicate media, invalid prompt references, and mismatched parameters that could cause generation failures. VIDEO GENERATION IMPROVEMENTS • Updated video requests to send the correct images, reference media, resolution, duration, and ratio for each model. • Improved text-to-video and image-to-video parameter handling to prevent cached text or settings from affecting a new task. • Models that do not support aspect ratio selection no longer receive an unnecessary ratio parameter. • Improved requests for multi-image, first-and-last-frame, and optional-audio models. • Continue checking task progress from generation results and history. • Added clearer background-generation guidance when leaving an active task page. RELIABILITY FIXES • Removed unavailable media from active reference positions. • Synchronized prompt references when associated media is deleted. • Prevented model settings, prompts, and references from carrying into a new task. • Improved media-duration reads, upload ordering, and request consistency. • Added clearer errors for invalid media or unsupported settings. • Fixed additional issues affecting uploads, task submission, and result viewing. Thank you for using Max Video. We continue improving model choices, reference controls, and generation quality to make the path from idea to finished work clearer, reliable, and flexible.