New This Month

Make targeted, instruction-based edits from one to four images with Bria's Fibo 2.0: background swaps, product restaging and campaign adaptations that keep the product true to the original. Add an optional mask to limit the edit to one area. Trained only on licensed data with clear commercial rights.

Generate campaign visuals, product and lifestyle scenes from a text prompt with Bria's Fibo 2.0, strongest on reasoning-heavy prompts with multiple subjects or spatial relationships. About one megapixel in 21 aspect ratios. Trained only on licensed data with clear commercial rights.

Animate an image into video with native audio using FLUX 3 Max. The image becomes the first frame; add an optional end image to set the last frame. 5 to 20 seconds at 720p or 1080p.

Generate video with native audio that passes through up to 10 pinned keyframe images in order using FLUX 3 Max. The first image opens the clip, the last closes it, and the rest fall evenly in between. 5 to 20 seconds at 720p or 1080p.

Generate video with native audio from a text prompt using FLUX 3 Max, a post-trained variant of FLUX 3 Video tuned for stronger prompt adherence, better visual quality and lower latency. 5 to 20 seconds at 720p or 1080p.

Edit and combine up to ten reference images from a caption of the result, with strong typography, up to native 4K and up to four images per run.

Fast, low-cost image generation from detailed prompts with strong typography and design, up to native 4K, with optional prompt expansion and up to four images per run.

Extend an image beyond its borders to a new aspect ratio. The original stays centered at full resolution and the new area is generated to match, with optional guidance for what fills it.

Remove the background and steer what stays with plain words, like "the person with the mat and the plant" or "only the chair". Keeps the original pixels with a real alpha channel; nothing is repainted.

Animate a photo of a person to speak or sing along to an audio track, with expressive, audio-matched movement. Optional prompt for scene, motion and camera. Audio 4 seconds to 5 minutes.

Transcribe up to two minutes of speech with optional word timestamps and speaker labels. Tuned for Japanese and East Asian languages; also English and Spanish.

Edit or combine up to eight reference images with text instructions, at up to about 2K. High reasoning plans the edit with segmentation and web references and fixes text.

Krea's third-generation image model at up to about 2K, with a quality (Berry) and a speed (Jam) variant and built-in prompt expansion.

Generate sound effects from a text description, up to 10 minutes long, with negative prompts and up to four variations per run.

Generate a synchronized soundtrack for a video and get it back as audio files, ready to mix. Keeps existing dialogue if you want, with text guidance and up to four variations.

Add a synchronized soundtrack to a video and get the video back with its new sound. Keeps existing dialogue if you want, with text guidance and up to four variations.

Natural, expressive speech in 44 languages with Cartesia Sonic 3.6, 3.5 or 3. Choose a library voice or one of your own voice clones, and steer emotion, speed and volume.

Clone a voice from a single 10-60 second recording. The clone becomes a custom voice you can pick in Cartesia Sonic Text to Speech.

Change a video's aspect ratio by generating new scenery around it, keeping the original timing and audio. Seven aspect ratios, 480p to 1080p, with optional guidance for what fills the new space. Source up to 15 seconds.

Generate new textures for an existing 3D model or a prior Meshy task from a text style prompt or a style image.

Animate a start frame into a 3 to 16 second clip with native audio, preserving the subject and composition of your image. Direct motion, camera cuts and spoken dialogue in the prompt. 540p up to 4K.

Build a 3 to 16 second scene from up to 12 reference images and 3 voice clips, keeping characters, products and voices consistent. Optional native audio with dialogue and sound effects, five aspect ratios, 540p up to 4K.

Change the lighting of a video from a lighting-sphere image: color, direction and shadow softness. Keeps the subjects, setting, motion, camera and audio. Source up to 15 seconds, 480p to 2K.

Fast, low-cost image animation from Kandinsky Lab: a 5-second clip with synchronized audio that keeps your start frame's composition and identity. Good for product shots and social content at scale.

Fast, low-cost text-to-video from Kandinsky Lab: smooth 5-second clips with synchronized audio, built for drafts and high-volume work. Understands English and Russian.

Animate a start frame into a cinematic 5-second clip with synchronized audio, preserving the subject, style and composition. Kandinsky Lab's flagship model.

Kandinsky Lab's flagship text-to-video model: cinematic 5-second clips with synchronized audio, camera moves, and strong adherence to long structured prompts. Understands English and Russian.

Upscale and restore video with sharper detail and stable, flicker-free frames. Works on generated and real footage; pairs with Kandinsky 6.0 drafts. Clips are capped at about 5 seconds.

Fast, low-cost video upscaling for previews and high-volume pipelines. Lower latency than Kandinsky 6.0 VSR, with slightly less detail. Clips are capped at about 5 seconds.

Continue an existing clip with 1 to 15 seconds of new footage you describe, faster and cheaper than H3 Max Extend. Optional audio references, 480p to 2K, and the choice of the whole extended video or only the new part.

Replace the people in a video with people from your photos, keeping the original motion, gestures, camera, cuts and soundtrack. One photo per new person, up to four. Source video 5 to 30 seconds, 768p or 1080p.

Google's high-efficiency image model. Generates from a prompt at up to 4K, can search the web for reference, and costs roughly a quarter of Nano-Banana Pro.

Edit images by describing the change. Upload up to 14 references to blend or restyle, at up to 4K. The same model as Nano-Banana 2.1, pointed at image editing.

Animate an image into a video with audio using Grok Imagine 1.5 Lite, a faster and lower-cost tier of Grok Imagine 1.5.

Generate videos with audio from a text prompt using Grok Imagine 1.5 Lite, a faster and lower-cost tier of Grok Imagine 1.5.

Edit, restyle, recolor, or combine up to 10 reference images with FLUX 3, served directly by Black Forest Labs. Place elements precisely with bounding boxes in the prompt.

Generate images from a text prompt with FLUX 3, served directly by Black Forest Labs. Supports up to 4K output and optional web/image grounding.

Drop a new scene into the middle of an existing video. Pick where it starts and where the original picks up again, describe what happens, and get back the whole video with the new footage spliced in. Only the new scene is billed.

Turn one product photo into a seamless loop that starts and ends on the same frame. The camera stays locked; pick gentle ambient motion or a full 360 degree spin. 5 to 15 seconds, up to 1080p.

Strip background noise out of a recording while keeping quiet and distant speech intact. Takes any audio or video file up to 30 minutes and returns loudness-normalised 48 kHz audio. The same model behind Clean Audio in the VEED editor.

Generate an image from a prompt with Ideogram 4.5, or supply up to 5 reference images (and an optional mask) to edit or restyle them.

Edit an image with Ideogram 4.5 while keeping every pixel you did not ask to change. Output matches the source image size exactly.