• Verbesserungen am Konversationsgedächtnis des Creative Assistant • Fehlerbehebungen und Optimierungen

Bild- und Videomodelle der nächsten Generation sind jetzt im Relaxed Supernet verfügbar, darunter Krea 2, Qwen Image Edit, Chroma, Flux, Z-Image, MiniMax H3 und weitere. Dieses Update bringt außerdem wichtige Verbesserungen bei Zuverlässigkeit und Performance für ein flüssigeres und stabileres Erlebnis.

Neue Supernet-Modelle — MiniMax FastH3 mit H3-LoRAs, Sound-to-Video, zweistufigen 1080p- / 2K-Presets, einer neuen Auswahl für die Videogröße, Krea 2 Identity Edit v0.3 und One-Click-Video mit FastH3. GPT Image 2.5 — Extra High- und Maximum-Qualität, transparente Hintergründe, Komprimierungseinstellungen und PNG-Bearbeitungsmasken. Prompts — modellspezifische Längenlimits mit Live-Zählern, die auch von Prompt Enhancer, Screenwriter und Songwriter berücksichtigt werden. Creative Assistant — DeepSeek V4 Flash, eine neue Option „Most Advanced Model“, verbindungssichere Antworten mit Send Again und ein Chat, der das aktuell angezeigte Bild versteht. Zuverlässigkeit — Behebungen für Ablehnungen bei fehlenden Uploads, verständlichere Fehlermeldungen, genauere Renderzeit-Schätzungen, Offline-Add to Queue und die Rückkehr des Spark-/USD-Schalters. Teste Sound-to-Video mit MiniMax FastH3!

Kritische Fehlerbehebungen, Stabilitätsverbesserungen und Leistungsoptimierungen für ein flüssigeres und zuverlässigeres Erlebnis.

Neue Modelle: • Seedance 2.5 — ein neues cinematisches Videomodell mit Unterstützung für bis zu 20 Bild-, 10 Video- und 10 Audio-Referenzdateien. • Wan 3 — ein neues Videomodell mit einer uneingeschränkten Enhanced-Variante. • MiniMax H3 Balanced — eine neue Qualitätsoption für H3-Videos. LoRAs — durchsuche und verwende LoRAs für Krea 2 und MiniMax H3 im neuen LoRA-Explorer, sowohl für Bilder als auch für Videos. Kreativassistent — LoRA-Unterstützung sowie Verbesserungen, Fehlerbehebungen und Optimierungen.

• RTX VSR-Bildskalierung — skaliere Bilder auf bis zu 16K hoch. • MiniMax Music 3 — erstelle vollständige Songs mit einer Länge von bis zu 5 Minuten, Gesang, Strophe-Refrain-Struktur und mehrsprachiger Unterstützung. • MiniMax H3 Video — enthält jetzt H3 Turbo für Reference-to-Video. • LTX 2.5 Turbo — ein neues Hochgeschwindigkeits-Videomodell. • Kreativassistent — Verbesserungen der Intelligenz. • Fehlerbehebungen und Optimierungen.

• MiniMax H3 Video: Eine neue Familie hochwertiger Videomodelle — Text-zu-Video, Bild-zu-Video, Erst-/Letztes-Frame-zu-Video und Referenz-zu-Video — jetzt an der Spitze der Video-Production-Auswahl. • Auto-format für H3: Aktiviere „Auto-format Prompt“ und Sogni formatiert deinen Prompt direkt vor der Generierung automatisch in das MiniMax-H3-Format um. • Krea 2 Bildbearbeitung: Neue Krea 2 Identity-Edit-Modelle für präzise, referenzgestützte Bildbearbeitung. • Aktualisierte Sampler & Scheduler: Überarbeitete Sampler- und Scheduler-Optionen für die neuesten Modelle. • Fehlerbehebungen und Optimierungen.

• Krea 2 Turbo ist jetzt der Standard — Supernet startet jetzt mit Krea 2 Turbo. • Steuerung der Prompt-Treue — Verbessere mit diesem neuen Regler die Prompt-Treue der Krea 2-Modelle. • Fehlerbehebungen und Optimierungen.

Four New Image Models Added Chroma1-HD, One Obsession v22, Dark Beast Krea 2, and Dark Beast Z-Image Turbo v9, with optimized defaults and gallery previews. Model Library & Generation Fixes Fixed missing Z-Image and Krea 2 model groups, generation requests getting stuck after content-filter rejections, and several error-message and popover issues across Manual Controls and Agent mode. Model Preset Improvements Added new scheduler options and corrected invalid recommended settings for Chroma v.46 Flash.

New Models • Added Krea 2 Turbo, Happy Horse 1.1, and Seedance 2.0 Mini. • Refreshed model defaults and library details. Happy Horse 1.1 • Full support for text-to-video, image-to-video, and reference-to-video with up to 9 reference images. • The Creative Assistant can now recommend and create with Happy Horse. Storyboard Video • Fixed Seedance storyboard clips so grid images and prompts stay properly connected. • Storyboards now support Happy Horse, including GPT Image 2 + Happy Horse workflows. Polish & Fixes • Chat notices are cleaner and no longer pile up during connection changes. • Fixed an Apple subscription issue that could block repurchasing after an expired transaction.

Unlimited Subscriptions Are Here New Sogni Pro and Unlimited plans with unlimited rendering, access to all Supernet models, and Premium Spark discounts. Manage Subscriptions In-App Subscribe, upgrade, downgrade, or reactivate without leaving the app. Access subscription controls from the Sogni menu. New Ace-Step v1.5 XL Audio Models Two new high-quality audio-generation models have joined the lineup. Creative Assistant Intelligence Enhancements Smarter assistance, improved model recommendations, and more reliable creative workflows.

New @ Mentions in Chat Type @ to quickly find and insert attached images, videos, or audio by nickname. Cleaner chat Nicknames Nickname editing now removes unsupported spaces and punctuation for more reliable mentions. Multi-Select Gallery Attachments You can now select multiple items from the in-app gallery and attach them to chat. Bug fixes and optimization

• Various bug fixes, optimizations, and UI refinements If you missed the previous update: New GPT Image 2 and Seedance 2.0 models. GPT Image 2 is now available with up to 16 context images and new Fast, HQ, and Pro quality modes. Seedance 2.0 adds advanced video creation with support for images, videos, and audio as reference media. Reference Media Panel Mix images, videos, and audio files as references for supported models and workflows, including GPT Image 2 and Seedance 2.0. Creative Assistant Upgrades The assistant now understands GPT Image 2, Seedance 2.0, Reference Media, quality controls, and improved model recommendations. Storyboard Grid Workflow Build multi-scene visual stories using GPT Image 2 and Seedance 2.0 through the Creative Assistant. Fixes and Improvements Improved job recovery, fixed recovered-job progress display, preserved generation settings from chat, and resolved several Model Explorer and UI issues.

New GPT Image 2 and Seedance 2.0 models. GPT Image 2 is now available with up to 16 context images and new Fast, HQ, and Pro quality modes. Seedance 2.0 adds advanced video creation with support for images, videos, and audio as reference media. Reference Media Panel Mix images, videos, and audio files as references for supported models and workflows, including GPT Image 2 and Seedance 2.0. Creative Assistant Upgrades The assistant now understands GPT Image 2, Seedance 2.0, Reference Media, quality controls, and improved model recommendations. Storyboard Grid Workflow Build multi-scene visual stories using GPT Image 2 and Seedance 2.0 through the Creative Assistant. Fixes and Improvements Improved job recovery, fixed recovered-job progress display, preserved generation settings from chat, and resolved several Model Explorer and UI issues.

Minor bug fixes and performance improvements.

This update brings major Creative Assistant improvements, a new video model, smarter media handling, and a smoother chat experience. • New LTX 2.3 Dev video model • New video analysis for attached videos • New short-film scene animation workflow • Improved Start & End Frame animation, including seamless loops • Smarter source image handling for attachments, generated results, and restored gallery items • Better model selection with support for shorthands like ltx23, wan22, flux2, and qwen-lightning • Smarter audio-aware routing for prompts with dialogue, music, or voice • Improved high-quality video model suggestions • Better suggestion chips, item nicknames, context tracking, and ambiguity handling • Chat fixes, including blank-screen fixes, smoother scrolling, and improved notices

Meet the new Sogni Assistant with creative superpowers! This major update brings a powerful AI creative assistant that can now help you with a wide range of creative tasks across image, video, and music workflows, with support for generation, editing, enhancement, animation, motion transfer, audio creation, prompt building, model selection, project guidance, and more. It can also manage multi-step workflows like creating ads from reference images, storyboards, short films, album covers, release packs, social media creatives, character design, restored photos, and music videos. Attach images, videos, and audio directly in the conversation, give them nicknames, and reference them naturally in your instructions. The assistant can analyze visuals, keep track of multiple assets across your project, and help you handle more advanced creative workflows with greater clarity and control. Music creation gets a big expansion too, with dozens of new style presets across new categories like Commercial & Advertising and Vintage & Era Styles, plus a greatly expanded Film, TV & Soundtrack section. Under the hood, this release also adds model quality ratings, deeper built-in knowledge about the Sogni platform, and many improvements that make the assistant more capable, aware, and reliable. Bug ​Fixes - ​Fixed paintover launch from workflows view. - Fixed size-preset check after model switch which could result in generation failures.

• Security enhancements add Keychain sharing across your devices and other Sogni apps for added convenience and credential backup. • Performance improvements and minor bug fixes.

• You can now set an end-frame reference image for i2v video without needing a start-frame image. • Faster, more reliable connections. • Improved item download reliability. • Improved jobs recovery after reconnecting. • Improved networking and auth stability.

Create soundtracks and songs with the new Music mode. Use Song Style prompts, add lyrics, explore style presets, generate in batches, and play results with waveform controls. You can also create videos from your music or use your tracks as soundtracks in Clip Mixer. Video creation also gets a big upgrade with LTX 2.3 support, including audio-to-video and image+audio-to-video workflows, new sample videos in Model Explorer, and an AI Screenwriter that helps turn simple ideas into richer video prompts. Video settings like FPS, frames, sampler, and scheduler are now handled more intelligently for each workflow. This update also adds more convenient features, fixes, and optimizations, including automatic aspect ratio matching when importing reference images and videos, a full rebuild of zoom, pan, and roll to fix memory issues, and improvements to Model Explorer with new Music filters and retry logic for failed samples. You’ll also find Sogni Restore in the workflows view, updated referral sharing, and fixes for PaintOver, Render Queue corruption, enhance workflow resets, size preset behavior, and other stability issues.

Big update focused on video creation, workflows, reference tools, and stability across iPhone and iPad. New Features • Create Videos: Create videos directly in Sogni • Clip Mixer: Full-featured video editor with timeline editing, clip trimming, reordering, export, waveform audio visualization, still-image clips, aspect ratio options, and multi-clip import • PaintOver Video: Bring your drawings to life in 4K video across supported video workflows, and save drawings to the gallery for reuse • Sound-to-Video (S2V): Generate videos driven by audio input (import from Files or Photos, or record directly) • Motion Transfer: Transfer motion from reference videos (WAN Lightning model) • First/Last Frame Controls: Define start and end frames for video generation, with video import to choose specific frames • Background Render Queue: Full background queue support with concurrent requests and result handling, plus a queue button and live status on the main view • Reference Tools Upgrades: Import gallery items as guide/reference/end-frame/motion references, plus Zoom / Pan / Roll controls for precise positioning • New Models: Added support for LTX2 video models and the new Z-Image base model • New Workflows View: Purpose-oriented quick access to tools, other Sogni apps, and helpful guidance Fixes & Improvements • Multiple bug fixes, stability improvements, and performance optimizations across queueing, Clip Mixer, gallery previews, PaintOver/reference tools, layouts, and overall iPhone/iPad experience

• New: Change Perspective. Re-render your image from a new camera angle, while keeping the subject and overall style consistent. • Fixes a bug that could affect the progress view layout. • Minor layout adjustments.

• New image models: – Qwen Image 2512 – Qwen Image 2512 Lightning • Fixes a bug that could cause recovered jobs to not load. • Bug fixes and optimization enhancements.

• New next-gen image models: – Z-Image Turbo – FLUX.2 [dev] – Qwen Image Edit 2511 – Qwen Image Edit 2511 Lightning • New Enhancer: Generative Filters, featuring one-click realism boosts, restoration, focus adjustments, color tuning, and more. • New Z-Image upscaler delivering consistent, high-quality images up to 2048 px. • Z-Image Turbo is now the default fast Supernet model. • Support for up to three context images when using FLUX.2 or Qwen Image Edit models. • Renamed the Model Explorer group “Flux” to “Next-Gen Diffusion”. • Added a new Model Explorer group: “Image Editing & Mixing”. • Added sample images, model descriptions, and in-app help content. • Fixed an issue where the prompt field appeared as a single line on launch. • Additional bug-fixes and optimization improvements. • Revised UI colors. • New app icon.

• Chroma models are now organized into two groups in the Model Explorer. • Refined Premium Spark purchase view. • More updates and fixes for Tahoe/iOS 26. • Improved app launch performance and reliability.