What's new
video-gen
- Added xAI's new video model, Grok Imagine Video 1.5. It creates videos from text or images up to 15 seconds long at 480p or 720p.
- The model generates audio alongside the video, including speech, music, sound effects, and ambience, while synchronizing lip movements with dialogue.
- Users particularly praise its speed, faithful preservation of the source image, and short-scene quality. Complex motion and voice generation can be inconsistent, so the model works best for concise clips with a clear prompt.
image-gen
- Grok Imagine 2 now generates higher-quality images at 1K resolution.
- The file size stays the same, but the picture looks cleaner and more detailed.
image-gen
- Creates photorealistic images with natural lighting, expressive textures, and cinematic composition.
- Supports a wide range of visual styles, from realistic photography to artistic and experimental imagery.
- Two-stage generation improves the detail of materials, faces, and small elements.
- Generates images at resolutions up to 2 MP.
- Operates without censorship and supports 18+ content.
video-gen
- Added MomentFlow v5.2. The new version generates videos up to 30 seconds instead of 15, handles longer narratives and scene transitions better, and maintains stronger visual consistency. Up to 30 images and 10 videos can be used as references.
- Reduced MomentFlow v5 Fast generation prices across all available resolutions, both with and without a video reference.
video-gen
- WildClips v4.0 is now available with native stereo audio: the model handles dialogue voiceovers especially well, synchronizes speech with on-screen action, and adds environmental sounds.
- Eleven languages are supported reliably: Russian, English, Chinese, Arabic, French, German, Italian, Japanese, Korean, Portuguese, and Spanish.
- Videos can be generated at up to 2K resolution and up to 15 seconds long.
- Text, first and last frames, images, video, and audio can be used as references (soon).
chat
- GPT models now cost half as much.
- DeepSeek-V4-Flash can now be paid for using your bonus balance.
image-gen
- MidJourney v8.1 and v7 are available again in the image generator.
- V8.1 supports 1K and 2K generation, while V7 is available in 1K.
- Both versions are listed under deprecated models.
image-gen
- V8.2 creates more expressive, bold, and refined images with a fresher aesthetic.
- Random low-quality results should occur significantly less often.
- While V8.1 focused on speed, prompt understanding, and detail, V8.2 primarily improves aesthetics and quality consistency.
- Problembo now lets you choose between 1K output and native 2K without additional upscaling.
image-genanime
- Richer image details.
- Vivid, high-contrast colors.
- More expressive and emotional character faces.
- An enhanced sense of volume and depth.
- More accurate adaptation to different painting styles.
image-editor
- Preserves the original shape and composition when changing colors, materials, backgrounds, or clothing, as well as when adding objects.
- Supports up to 5 references for consistent characters, product shots, and assembling scenes from different sources.
- Separates the result into editable layers and automatically reconstructs the background after objects are removed or moved.
Page 2 of 10