@blizaine

Maestro v1.3.0 is out: SCAIL-2 character animation, 100% LOCAL, FREE & EASY!

(NEW) "Recast": swap anyone in a video for your own character. Drop a clip, type who to replace ("the woman", "man in red"), drop a character image, hit generate. Masking is fully automatic. SCAIL-2 Fast: bundled distill runs 6 steps with no CFG, about 13x faster than stock. Looks great! (used in video below). New gallery trick: scrub any video to the perfect frame & send it straight to your image gen references. Free, open source, runs on your GPU. One-click install via Pinokio.

2d ago

Maestro v2.6.0 is out! Live Preview & Reference Mode long videos

Maestro v2.6.0 is out! MiniMax H3 Reference Mode now supports long videos (like this stock footage) & Maestro auto-splits across seamless windows. Combine w/ Character Swap LoRA; keep the background true, or add effects.& x20; Maestro v2.6.0 Updates: • Live generation previews: Watch motion develop while rendering, with looping Tiny VAE video, pause/resume, and mobile playback. Now enabled by default. Can be turned off in settings. • Bring your own H3 checkpoints: Download compatible community models directly from the Model Browser, via URL or CivitAI, including INT4 and GGUF, with verified re

13d ago

Maestro v2.4.0 - Lots of good stuff! See notes!

I knew this guy was up to no good. Also, Maestro v2.4 is out. You too can make ridiculous things like this locally for free! Maestro v2.4: • Fullscreen gallery: swipe navigation, favorites, tap-to-pause, remembered sound, smoother transitions and pinch-zoom. Social video style (TikTok/Reels) • Before/after image comparison with a draggable divider. • Mobile improvements: reliable video posters, better viewport positioning and collapsible landscape sidecar. • Gallery input routing: send images, captured frames and videos into the active Studio or Director mode. • Qwen Image 2.1: CivitAI LoRA fi

17d ago

Maestro v2.3 is out! Qwen Image 2.1, YuE2 Training w/ Vocals, 30s H3

• Qwen Image 2.1 generation & editing • 30-second H3 clips (Experimental) • YuE2 - Auto & Guided music LoRA training w/ vocals support (Experimental) • YuE2 - instrumental support

21d ago

Maestro v2.2.0! YuE2 Generation & Training! Major Prompt Enhancement Improvements!

You need to try YuE2 NOW! It’s Suno v5 level, open weights, low VRAM, FREE, FAST music gen & can even be trained in Maestro v2.2.0! Also, MAJOR v2.2.0 AI prompt improvements! GPT-6 ASTRA trained Maestro’s local agents & made them awesome! Full Maestro v2.2.0 Release Notes: Unified AI Prompt Enhancement, YuE2 music, and more control over Director productions •One Enhance workflow: develops short ideas and adapts detailed scripts to the selected model. Enhance now, or let a queued job enhance immediately before generation. An optional Use by default setting remembers Enhance on generation; saved

29d ago

Maestro v2.1 is live!

Maestro v2.1 is live! 🎬 • Viggle Animate Support! • Streamlined UI & Menus • Save & share characters w/ RefMod support • H3 Voice Audio: generate long dialogue w/ SFX • H3 Face Refinement • DLSS 5 upscaling Maestro = FREE, OPEN, LOCAL!

Maestro v2.0 is out! Its got stuff and things!

Maestro v2.0 is out! & it’s loaded with new features! Maestro is a free, open, local, unrestricted AI media studio that’s loaded w/ the latest SOTA models & can autonomously gen films & music videos. v2.0.0 (2026-09-03) A complete create-to-edit workflow - Added the new full-screen Editor Mode with multi-track video, audio, and title editing; 21:9 canvases; canvas transforms and snap guides; transitions, speed, opacity, volume, fonts, undo/redo, project history, and hardware-aware export. - Director productions can be opened as editable timelines with separate shot clips and the co

🎬 Maestro v1.9.0 is live! - Universal Director & Studio generation queue + tons of bug fixes!

Highlights: • Universal Director & Studio generation queue • Queue multiple projects for unattended overnight rendering • Persistent Director projects with full recovery and Load Settings • Faster, more memory-efficient MiniMax Music3 generation Full Release Notes: v1.9.0 (2026-08-19) Universal queue and Director recovery - Added one global Studio + Director queue with a compact top-bar popover, live count badge, ordering controls, removal, pause, and start controls. - Studio's split Generate button can now hold complete jobs without starting them, so several prompts can be prepared before

Maestro v1.8.5 with MiniMax-Music3 & other fixes and improvements!

Haha! This is crazy. Maestro v1.8.5 w/ MiniMax-Music3 just dropped. As a test, I prompted "Wheels on the Bus as a death metal song by the band Nursery Crimes." But it's Maestro, so I told it to auto-make the entire video using MiniMax-H3 for video.🤯 Free, Local & Open! 🔊🎵📽️ v1.8.5 (2026-08-14) MiniMax-Music3 - Added native local MiniMax-Music3 generation for complete stereo songs from structured music direction and lyrics, with selectable 5-second to 5-minute runtimes and a two-minute default. - Added a duration-aware AI song writer that scales lyrics, sections, arrangement, transition

Maestro v1.8.0 is Live! MiniMax-H3 Sol Engine Speed ups & LTX-2.5 Support!

v1.8.0 (2026-08-13) LTX-2.5 and next-generation LTX workflows - Added native local LTX-2.5 with synchronized audio. Distilled is enabled by default, while Dev and NVFP4 variants can be enabled in Settings. - Added the official Distilled 8-step base pass, learned latent upscaling, and 3-step full-resolution refinement, with persistent model reuse for faster follow-up generations. - Added first and last frames, timed frame injection, audio-driven video, control-video audio, native audio, and compatible LTX-2/2.3 LoRAs from the existing shared library. - Added LTX-2.5 to compatible Director Music

Maestro v.1.7.0 - BIG UPDATE! Long multi-clip videos in FL & REF models

v1.7.0 (2026-08-10) MiniMax H3 native multi-window generation - Added native multi-window continuation to both First / Last and Omni, carrying recent motion and synchronized stereo audio into each following window for smoother transitions. - Added shared Multi-window controls with total duration, editable per-window prompts, optional AI planning, and independent hard-cut sequences. - Improved H3 sequence planning so actions, dialogue, camera cuts, sound, and story events advance across windows instead of repeating or finishing in the first clip. - Added exact-duration assembly, model-aware ove

Maestro v1.6.5 - MiniMax H3 Fixes & Improvements!

v1.6.5 (2026-08-08) MiniMax H3 performance and lower-VRAM support - H3 Turbo now works with the recommended Pruned 20B models as well as the optional Full 33B models. - Turbo now starts at six steps and LoRA strength 0.50, while keeping the LoRA visible and adjustable in Advanced settings. - Reworked H3 model residency, activation chunking, and VRAM budgeting to reduce step-zero out-of-memory failures and excessive CPU offloading. - Added resolution- and GPU-aware First / Last window recommendations, with clear warnings and a manual override for experimental combinations. - Added an optional e

Maestro v1.6.0 is out! Major MiniMax H3 Improvements!

v1.6.0 (2026-08-06) MiniMax H3 Omni Reference - Added MiniMax H3 Omni for generating new video and synchronized audio from ordered image, video, voice, motion, and sound references. - References can be reordered, labeled with their intended role, and used for identity, appearance, scene, motion, voice, performance, ambience, or music conditioning. - Added both recommended Pruned 20B and optional Full 33B Omni models. - Added Match Output reference preparation for consumer GPUs and an optional Maximum Detail mode for higher-memory systems. - Improved reference-video memory use with output-aware

Maestro v1.5.5 with MiniMax H3 is out!

v1.5.5 (2026-08-04) MiniMax H3 local audio-video generation - Added native local MiniMax H3 Base FL2VA support with text-to-video, image-to-video, and first/last-frame video generation. - H3 generates synchronized 32 kHz stereo audio together with the video instead of requiring a separate audio pass. - Added approximately 5-15 second generation at 24 FPS with landscape, portrait, square, native 768p, and lower-VRAM resolution options. - Added automatic, revision-pinned provisioning for the compact scaled-FP8 transformer, NVFP4 Qwen3-VL conditioner, video VAE, audio VAE, tokenizer, and processo

Maestro v1.5.0 is out! - Rebuilt Recast & Added Repaint (SCAIL-2), Major LTX Outpaint Improvements, Krea 2 image edit & more!

v1.5.0 (2026-08-02) SCAIL-2 Recast and multi-character replacement - Rebuilt Recast around SCAIL-2's native replacement conditioning for substantially stronger identity transfer and motion tracking. - Added color-mapped character cards for replacing up to five people in one run. - Added camera-shot detection and per-shot processing so characters remain correctly mapped when a video cuts between close-ups, wide shots, and group shots. - Improved two-person and multi-person shots by conditioning each shot only on the characters visible in it. - Added automatic reacquisition when a person first a

Maestro v1.2.0 is out 🎨🎵

Two big additions this release: - Light themes. Every Maestro theme now has a daylight version. Pick your style (Golden Hour, Classic, or Onyx) and set the mode to Dark, Light, or Auto. Auto follows your system's appearance and switches live, so scheduled dark mode at sunset just works. - ACE-Step v1.5 XL SFT LM_4B 4B is now the default music model. It is absolutely the BEST open weights music generation model available. This is the quality-focused variant of the XL 4B music model, the one that runs full classifier-free guidance instead of the fast distilled schedule. Until now, it meant cloni

Maestro update v1.1.3

New features - Linked Model Folders: point Maestro at your other installs (like Wan2GP) and reuse their checkpoints and LoRAs instead of re-downloading hundreds of GB. One-click scanning of your Pinokio apps, strictly read-only, and AI LoRA guides work on linked LoRAs too. (see image below) - Krea 2 image models (Raw and Turbo). - 10Eros v1.4 with the author's abliterated text encoder and per-stage LoRA strengths. - Reference Pipeline toggle for 10Eros models (on by default): runs the model author's published ComfyUI workflow config, with hand-tuned sigmas, per-step CFG and STG, and rectified-

3mo ago

Say hello to Maestro!

Finally! Say hello to Maestro! Describe a music video or short film, and it makes the whole thing: original song, plans every shot, screenplays, generates the clips, & cuts it all together. 100% Local. No cloud, no subscription. FREE. One-click install via Pinokio. Director mode: give it a prompt (or your own track), and a local LLM writes the screenplay + lyrics, analyzes the beat, plans shots with character consistency, runs the full multi-clip generation, and edits the result. Music videos and short films. Want hands-on control? Studio Mode exposes every model and knob intuitively: LTX-