August 26th, 2026
New
Your videos just got twice as long. Wan 3.0 pushes the single-generation limit from 15 seconds to a full 30 β no stitching required. It's live now across Text to Video, Image to Video, and Video to Video.
Strengths:
Up to 30 seconds in a single generation β enough room for a full scene, not just one beat of motion
Follows a multi-shot script, not just a single action β describe several shots by timestamp and the model handles the transitions between them itself
Writes its own soundtrack β dialogue, foley, and music synced to the video when you turn on Generate Audio
Holds multiple references at once β images, video, and audio together steer identity, motion, and sound
480p, 720p, or 1080p β pick the resolution that fits the job, from quick drafts to final renders
Long scenes with multiple characters and continuous action that used to take several linked clips now fit in a single, well-written prompt.
New to Wan 3.0? The Mastering Wan 3.0 guide covers how to size your prompt to the clip length β the full course inside the app goes deeper.
August 20th, 2026
AI Prompt Assistant is a new chat panel built into the tool β open it with one click, describe your shot in plain language, and get back a structured, ready-to-use prompt in English.
What's new:
Available in three tools β Text to Video, Image to Video, and Video to Video
Any language in, English out β describe your idea however comes naturally, the assistant writes the prompt in English
Settings-aware β automatically tailored to your current duration, resolution, and style, no need to repeat them
References your uploads β mention an uploaded image by its tag (e.g. [img1]) and the assistant builds the prompt around it
Free to use β chatting never spends credits, only generating the video does
August 14th, 2026
LoRAs AI Effects is a brand-new hub β one library for every LoRA style we support, built on Wan 2.2, Wan 2.7, and Flux Klein.
What's new:
Video or image, your call β browse Video LoRAs (Wan 2.2, Wan 2.7) and Image LoRAs (Flux Klein) separately, or see everything at once
Sort and filter however you work β Trending, New, by clip length (5s/10s/15s), or search by name
A growing set of scenes and looks β there's something here for everyone
π Favorite as you go β heart any style once, and it's one click away every time after that.
π Try it: LoRAs AI Effects
July 30th, 2026
Companion is a chat that goes beyond conversation β it writes expert prompts for you and generates photos and videos right inside the chat.
What makes it different:
Say it in plain words β describe what you want, Companion turns it into an expert prompt
Photos & videos land right in the conversation β no need to jump to a separate tool
Generation uses your existing credits β billed at normal rates; the subscription itself covers unlimited text messages, not media generation
Companion can hold a conversation on any topic π₯
π Founding Member offer β the first 100 subscribers get Companion for $19.99/month with unlimited messages. Seats are limited and going fast.
Coming soon: roleplay that plays along, and a Companion that remembers your name, inside jokes, and where your last conversation left off.
π Try it: Companion
July 29th, 2026
Seedream 5 Pro just landed in three tools at once β Reference-to-Image, Image Editor, and Text to Image β plus a model swap in PhotoShoot.
It's a real step up in photorealism over earlier versions β more accurate lighting, materials, and skin texture β so results read less like typical AI output and closer to an actual photo.
What makes it different:
Scene or character swap, pick your mode β in Reference-to-Image, place a face's identity into any reference (Scene mode), or swap up to 4 characters directly into a source scene (Character swap mode)
Keep Pose β send a reference straight through with a fixed prompt, so pose, composition, outfit, body, and background all carry over from the reference β only the identity changes
1K or 2K, your call β every tool lets you pick faster, cheaper 1K or larger, more detailed 2K
PhotoShoot got the upgrade too β the model's now Seedream 5 Pro, for sharper, higher-quality photoshoots of your characters
π Try it: Reference-to-Image Β· Image Editor Β· Text to Image
July 29th, 2026
We've added SeedVR2 to Image Upscaler β a powerful upscaler from ByteDance. Most AI upscalers just enlarge an image; SeedVR2 also cleans up noise and blur while it does, so results come out sharper without extra artifacts.
What's new:
SeedVR2 3B Base β adaptive upscale with selectable detail, up to 8K Β· 1 credit per image
SeedVR2 7B Premium β up to 6K Β· 36MP Β· 2 credits per image
Synthetic detail control β dial in how much extra detail SeedVR2 Premium adds while upscaling
β‘ One pass, not several β SeedVR2 handles cleanup and upscaling together instead of running them separately, which is why it holds up especially well on faces, hair, skin, and fine text.
π Try it: Image Upscaler
July 29th, 2026
We've added the Fast β MuleRouter model to the Head & Face Swap tool, built for quick turnaround.
What's new:
Face or full head, your choice β swap just the face, or the entire head, whichever the shot calls for
Our fastest option yet β optimized purely for speed, with output delivered as-is
π Try it: Head & Face Swap
July 28th, 2026
New
This one's been requested for a long time β the ZenCreator interface is now fully translated, not just parts of it.
What's new:
Every screen, fully translated β this isn't a partial rollout, the whole interface works in your language
6 languages live today: English, EspaΓ±ol, ζ₯ζ¬θͺ, Deutsch, FranΓ§ais, Italiano
More on the way β this is just the start, the list will keep growing
Pick it once, done β choose your language from the switcher in the top bar, and it stays selected
July 15th, 2026
Wan 2.7 + LoRAs is here β a full new set of style templates built on our latest and best-performing model.
What's new:
Clips up to 15 seconds β choose 5s, 10s, or 15s, enough room for a full scene when you need it
Audio generated right alongside the video β no separate step, no silent output
A wide range of sex poses and scenarios β the new LoRA styles support many different movements and setups
720p and 1080p output β including full 1080p for sharper, more detailed results
Pick a LoRA style, upload your reference photo, and get a full 15-second clip with sound β your reference brought to life with that template's movement, in one generation.
π Try it: Image to Video
July 9th, 2026
Seedance 2.0 just landed in three tools at once β sharper motion, smarter scene understanding, sound generated right alongside the video, and full NSFW support across all three.
What makes it different:
Understands the whole scene, not just a picture β subject, action, environment, and camera all come together into one coherent clip, with noticeably more stable motion than before
Two distinct input modes β animate a clip with a first/last frame, or try the new Reference mode: build a full clip from up to 9 image references and 3 audio clips
Flexible clip length, up to 15 seconds β more room to build out a scene, not just short bursts
Built-in audio generation β sound and motion land in sync without a separate step
Precise video editing β in Video-to-Video, swap a costume, background, or lighting while keeping the original movement and timing untouched
β¨ Try Seedance 2.0: Text-to-Video Β· Image-to-Video Β· Video-to-Video