New models and features in AI image, video, voice and music generation, from Midjourney, Sora, Veo, Suno and more.
ElevenLabs just rolled out a major overhaul to Studio, its long-form audio narration tool. The update adds word-level regeneration to fix flubs without burning extra credits, a lock feature to secure finished paragraphs, and multi-voice support within a single paragraph. It makes producing full-length audiobooks and dialogue-heavy scripts far less tedious.
ElevenLabs just partnered with the Jerry Garcia Estate to bring the late Grateful Dead frontman's AI-recreated voice to its ElevenReader app. The model can read audiobooks, articles, and PDFs across 32 languages, with plans to expand into future narrated documentaries. While it only reads text rather than sings, it marks another step in turning iconic celebrity legacies into interactive AI assets.
ByteDance just unveiled X-Portrait 2, an AI model that transforms a single still portrait into a remarkably nuanced video performance mimicking famous movie scenes. Instead of relying on dot-tracking, it captures fluid facial muscle movements directly. The company is keeping the tech private over deepfake concerns.
Black Forest Labs just rolled out Ultra and Raw modes for FLUX1.1 [pro]. Ultra mode delivers 4x higher resolution up to 4MP in just 10 seconds for $0.06 per image, while Raw mode ditches the synthetic AI sheen for natural, candid photography aesthetics. Both are now available via their API.
Kuaishou's Kling AI just started testing a new Custom Models feature that tackles character consistency in video generation. Users can train personal characters using 10 to 30 short video clips and tag them directly in prompts to keep faces stable across angles. Why it matters: persistent character identity has long been the hardest bottleneck holding back AI filmmaking.
Runway just rolled out advanced camera controls for its Gen-3 Alpha Turbo video model, giving creators direct control over movement speed and direction. Users can combine pans, zooms, and tracking shots instead of relying purely on prompt luck to steer the camera. The feature is available now for all Gen-3 Alpha Turbo users.
Suno just launched 'Personas,' a new feature that lets users capture and reuse the exact vocals, style, and vibe of an existing track across new generations. Creators can save their custom singer as private or share it publicly. It is currently in early-access beta for Pro and Premier subscribers with 200 free tracks, costing 10 credits each afterward.
Blockade Labs just launched Blendbox Alpha, a new creative tool designed to give artists precise control over AI-generated visuals without wrestling with long prompts or random outputs. Why it matters: It shifts AI image creation away from prompt lottery toward direct, hands-on visual shaping.
D-ID just released two new AI avatar models, Express and Premium+, built for real-time interaction. While Express needs only one minute of video to replicate head movements, the flagship Premium+ adds hands and torso gestures for more natural conversations alongside a new enterprise marketing suite.
Recraft just launched Recraft V3, claiming the #1 spot on Artificial Analysis's text-to-image leaderboard with an 1172 ELO rating, surpassing Midjourney and OpenAI. The new model lets creators place and scale exact text within designs, handles long-form copy, and natively generates vector graphics. It is available immediately for free and paid tiers across web, mobile, and API.
A mysterious new AI image generator called Red Panda just claimed the top spot on Artificial Analysis's Text-to-Image Arena leaderboard. It boasts an Elo score of 1172 and a 72% win rate, beating out Midjourney v6, DALL-E 3, and FLUX 1.1 [pro]. The creator remains unknown, but it is likely an unreleased model undergoing blind testing before launch.
Universal Music Group just partnered with AI startup KLAY to build an ethical foundational music model called KLayMM. KLAY, led by music industry and DeepMind veterans, promises not to compete with artists' catalogs while ensuring accurate attribution. It is UMG's latest push to legitimize commercial AI music on its own terms.
Meta just unveiled Movie Gen, a powerhouse suite of media models featuring a 30B video generator and a 13B audio model. It can create high-definition footage, sync ambient sound and foley, edit specific video elements with text, and turn single portraits into personalized clips. It is currently being tested with creative professionals ahead of a potential release.
DreamWorld just announced a Steam playtest featuring built-in AI text-to-3D asset generation powered by Meshy. Players can type prompts to create custom 3D models directly inside the infinite sandbox MMO alongside advanced building tools. Why it matters: Bringing generative 3D straight into live multiplayer worlds could make user-generated content vastly faster and easier.
Black Forest Labs just launched Flux 1.1 Pro, an updated image generation model claiming six times faster generation speeds alongside improved visual quality and prompt adherence. The startup also launched a beta API, letting developers integrate the Flux model suite into their apps via its platform or partners like Replicate.
ByteDance just unveiled two powerful video generators, PixelDance and Seaweed, stepping up its challenge to OpenAI's Sora. PixelDance generates 10-second clips while keeping characters and scenes consistent across varying camera angles via simple text prompts, while Seaweed scales consistent environments up to two minutes. Both are currently in private beta, targeting one of AI filmmaking's biggest bottlenecks.
Pinterest just launched generative AI features inside its Performance+ ad suite, letting advertisers turn plain white product shots into lifestyle imagery. In early tests with Walgreens, the AI backgrounds delivered a 55% jump in clickthrough rates while cutting campaign setup inputs in half.
Pika Labs just launched Pika 1.5, introducing Pikaffects that let users generate surreal, physics-defying video clips directly from text or images. The update also improves cinematic motion control. While rivals race for pure realism, Pika is leaning hard into creative, mind-bending effects.
Luma AI just launched Hyperfast video generation, boosting Dream Machine v1.6 inference speeds by 10x to deliver full-quality clips in under 20 seconds. The company noted it didn't rely on lighter "turbo" or distilled models, keeping output quality intact. It's live today for all paid subscribers and API customers.
Civitai just unveiled SPINE, a unified creative workflow integrating leading external generative tools like Kling AI, Udio, and ElevenLabs into a single workspace. Creators can now string together multi-model media pipelines without tab-juggling, paying via Civitai's Buzz credit system.
