Sora 2 & GPT Image 1 are now available on fal
We’re thrilled to announce that Sora 2 and GPT Image 1 are now available on fal.ai. These groundbreaking generative AI models push the boundaries of what’s possible in video and image creation. Whether you’re building creative tools, social media content, or interactive experiences, these new releases redefine realism, speed, and creative control.
🎬 Sora 2
Sora 2 arrives as OpenAI’s most advanced video generation model, and fal gives developers and creators the easiest, fastest way to run it, without watermarks and with full creative freedom. Designed for high-fidelity storytelling and production-grade quality, Sora 2 on fal unlocks the full potential of generative video for creative teams and developers.
Key Model Strengths
1. State-of-the-Art Model
Sora 2 represents a monumental step forward in AI-driven video creation. Built on a cutting-edge transformer architecture optimized for long-range temporal coherence, it produces footage that feels cinematic, detailed, and deeply realistic. The model captures lighting nuances, texture consistency, and camera-like motion with incredible precision, creating visuals that can easily pass as real-world footage.
2. Native Audio Generatio with hyper-realistic Voices
Sora 2 introduces native audio generation, allowing sound to be an integral part of every video. It can create synchronized, context-aware audio tracks, from realistic songs to environmental noise to fully voiced dialogue. The voices it generates are expressive, emotionally rich, and realistically timed with lip movements or actions on screen. For creators, this means no more post-production audio syncing; instead, the visuals and sound emerge together as one cohesive, cinematic experience.
3. Realistic Cuts and Multi-Scene Generation
Sora 2 isn’t limited to a single shot, it understands narrative structure. The model can create multi-scene videos with natural transitions, camera changes, and pacing that mimics professional editing. Whether it’s switching from a close-up to a wide aerial shot or moving through time and location, Sora 2 maintains continuity in lighting, color grading, and character identity. The result is a fluid, film-like output that can tell cohesive stories in one generation. For example, "Tom's Furniture" ad created bellow was generated from a single prompt, and includes multiple coherent cuts that form a full professional ad. For filmmakers, ad creators, and game designers, this introduces a new level of narrative expressiveness that was previously impossible in AI-generated video.
4. Optimized for Social Media and Creator Workflows
With Sora 2 on fal, content creation for social platforms becomes effortless. The model supports optimized aspect ratios and loop-friendly motion, ensuring videos are ready to post on TikTok, Instagram, or YouTube Shorts right out of the box. It can generate dynamic vertical clips, branded transitions, or cinematic reels in seconds, with consistent quality across formats.
5. Remixing and Video Continuation
Sora 2 introduces powerful remixing capabilities, allowing creators to build upon existing videos while preserving key elements like scene composition, character identity, and motion flow. By uploading a reference clip, you can guide the model to generate new sequences that match the original’s style, pacing, and environment — extending scenes, introducing variations, or creating entirely new storylines that feel cohesive.
GPT Image 1
GPT Image 1 is OpenAI’s most advanced image generation model, now available on fal, giving creators and developers access to hyper-realistic image synthesis, multi-image editing, and precise text generation within images.
1. Hyper-Realistic Results
GPT Image 1 delivers photorealism at an entirely new level. Every pixel carries a sense of depth, texture, and light realism that feels natural and cinematic. From fashion photography and product renders to architecture and landscapes, the model captures shadows, reflections, and fine-grain details with uncanny precision.
2. Multi-Image Editing and Composition
With multi-image editing, GPT Image 1 lets you merge, transform, or reimagine multiple inputs into one coherent visual. You can combine different reference images, preserve brand elements, or make detailed localized edits while maintaining consistent lighting and perspective. This opens powerful workflows for designers and creators, from refining concept art to building product catalogs or campaign visuals.
3. Accurate and Natural Text Generation
Unlike earlier image models, GPT Image 1 can generate text inside images with clarity and precision. It understands typography, spacing, and composition, producing legible, stylistically appropriate text that fits naturally within the scene. This makes it ideal for marketing visuals, posters, ad creatives, or any context where design meets language.
Getting Started with Sora 2 & GPT Image 1
The easiest way to explore Sora 2's and GPT Image 1's capabilities is through Fal's Playground, where you can experiment with prompts and see immediate results. A detailed guide on how to integrate the models into your platform is available in our API documentation.
Sora 2 endpoints:
GPT Image endpoints:
Stay tuned to our Reddit, blog, Twitter, or Discord for the latest updates on generative media and the new model releases!