The latest AI news and updates of fal — Generative media infrastructure platform for serverless inference of image, video, and audio models at scale. Covering fal's latest product updates, launches, and company news from the past 90 days.
falSep 10fal has made ByteDance's Dreamina Seedance 2.0 available to US companies, with inference hosted entirely on US infrastructure. The model supports native audio-video generation and multimodal control across text, images, video, and audio for production-grade workflows. Access is available through the fal sales team.
falSep 9fal launched Multi-angle for its H3 Max video model, enabling precise camera movement via horizontal and vertical degree settings. The system renders continuous shots from a single viewpoint in under 3 seconds, maintaining consistent subject identity and scene geometry throughout the motion. Promotional launch pricing is available at 75% off through September 14.
falSep 9fal now hosts Bria AI's Ad Delayer, a model that converts flat ad images into structured, editable layers. The system recovers headlines as SVG text for structural localization and organizes assets into typed, ordered layers. The model is available via the fal API, with generation priced at 0.30 dollars per request.
falSep 8fal now hosts OpenAI's GPT Image 2.5, featuring instruction-scoped editing that preserves composition and background. The model maintains subject identity across new settings and styles, ensuring consistency during long editing sessions. It is available via Flare and Sunburst endpoints for text-to-image and image-editing tasks, supporting high-quality generation and transparent backgrounds.
falSep 8fal launched Phota Identity v2 by PhotaLabs, a system for creating persistent identity profiles from as few as 5 photos. The model generates a reusable identity in about a minute, enabling consistent subject likeness across expressions, poses, and scenes. The service is available via four API endpoints, with pricing starting at $0.09 per 1K image generation.
falSep 8fal now supports native, real-time Full HD (1080p) video generation for its H3 Max and H3 Max Turbo models. Both models produce 5 seconds of video in under 5 seconds of generation time, delivering increased detail in every frame. This capability is available across all H3 Max text-to-video, image-to-video, and reference-to-video endpoints.
falSep 6fal released general availability for Reference-to-Video on its H3 Max model, achieving an RTF of 0.876. This update enables real-time generation with up to four references, delivering up to 2X faster performance. The model now features improved semantic alignment between prompts and references, alongside stronger character consistency through continued reinforcement learning.
falSep 3fal released the H3 Max Director API, a video model that generates a single continuous, steerable stream while maintaining character and scene continuity. Unlike models that stitch discrete clips, this system allows live prompts to evolve action in real time. Generations are 75% off for the first two weeks, with pricing starting at 0.02 dollars per second.
falSep 2fal added Bria AI's Fibo Edit 1.5, a multi-reference image editing model. The system accepts up to four reference images, allowing the identity of products, garments, or characters to persist through edits. It supports natural language and structured JSON instructions, targeting catalog and advertising workflows like virtual try-on and background replacement. Pricing is set at 0.04 dollars per image.
falSep 2fal released a preview of H3 Max Turbo, a video generation model that runs at twice the speed and half the cost of the original H3 Max. It maintains 97% of the original model's quality and is available at a promotional price of $0.01 per second for 768p output for the next two weeks.
falSep 1fal relaunched its fal.live platform, which now runs continuous AI-generated broadcasts directed by viewer-upvoted LLM prompts. The system utilizes fal’s faster-than-real-time H3 Max video model to generate content in response to incoming audience instructions. Viewers can submit ideas for new channels or apply to stream their own content on the platform.
falAug 31fal launched an early preview of Reference-to-Video for its MiniMax H3 Max model. The feature achieves up to RTF=1 on 768p resolution, or RTF=1.5 when using references. fal aims to double the speed and reduce costs for this capability within the week while maintaining the model's performance.
falAug 31fal launched fal.live, an interactive platform for infinite AI-generated broadcasts. The system generates continuous video content that responds to text prompts, allowing the stream narrative to evolve based on viewer input. This platform provides a broadcast experience where the video output updates in real time according to incoming instructions.
falAug 30fal launched H3 Max Live, an infinite broadcast where video generates faster than real time and scenes are directed by chat prompts. The system runs on an experimental long-form H3 Max checkpoint with native infinite continuity, allowing the model to maintain thread consistency across scenes. This long-form capability will be available via the fal API next week.
falAug 27fal now hosts Meta's Muse Image model, an agentic system that plans compositions, uses tools, and self-corrects to improve output fidelity. The integration provides text-to-image and image-editing endpoints, with generation priced at $0.01 per image. The model renders fine details like text and QR codes with high accuracy.
falAug 27fal Research launched H3 Max, a post-trained video model built on MiniMax's open-weight H3. It generates 5-second 720p clips in under 3 seconds, providing roughly 35x the throughput of the base endpoint. The model ranks #1 in Image-to-Video and #3 in Text-to-Video on Artificial Analysis leaderboards. It is available now with 50% off pricing through September 1.
falAug 26fal has integrated Recraft V4 Styles, enabling style-consistent image and vector generation from a single reference image. The platform offers a Precise mode for meticulous matching and a Flexible mode for general style adherence. Each request costs $0.005, providing a reusable style ID for consistent output across multiple generations.
falAug 25fal has integrated Alibaba’s Wan 3.0 video model, enabling native 30-second clips generated in a single pass without stitching. The model supports omni-reference inputs including text, images, audio, video, web pages, and documents. It delivers reality-grade rendering with improved motion, with pricing starting at $0.05 per second for 480p output.
falAug 22fal launched the Black Forest Labs FLUX Video Upscale API, supporting video upscaling to 1080p, 2K, or 4K. The service offers a Precise mode for identity preservation and a Creative mode for detail enhancement, with optional prompt-based steering and source audio retention. Pricing starts at 0.14 dollars per second for 1080p output in Precise mode.
falAug 22fal launched Hi3D v3.0, an image-to-3D model featuring 2048³ voxel resolution. The model generates high-precision geometry with clean topology, preserving thin structures and dense surface ornament from a single image. It is available via two new API endpoints, with generation costs starting at $2.10 per request for 2048³ quality.
falAug 22fal now hosts 16 Topaz Labs enhancement endpoints for video and images, including Starlight Precise 2.6, Wonder 3.5, Astra 2, Bloom 2, and Hyperion 2.5. The integration supports upscaling, restoration, denoising, interpolation, and SDR-to-HDR conversion via a single API. Starlight generative video upscaling starts at 1.20 dollars per 10 seconds for 1080p output.
falAug 15fal is opening applications for a creative partner cohort, providing selected filmmakers, studios, and agencies with $1,000 in fal Agent credits. The program supports production workflows requiring character and style consistency across generative media models. Applicants receive credits to build narrative projects, advertising campaigns, and editorial assets using the agent’s persistent memory and cross-session continuity features.
falAug 12fal launched Meshy V7, enabling the conversion of single or multiple images into textured, PBR-ready 3D models. The integration supports production-grade materials, clean topology, and optional auto-rigging with ready-to-use animation in a single pass. The model is available now via fal’s API, with pricing starting at $0.80 for base meshes and $1.20 for fully textured outputs.
falAug 11fal launched Lightricks' LTX 2.5, an open-source audio-video model featuring Diffusion Fidelity Rendering for improved compute in complex scenes. The platform provides six endpoints across Fast and Pro tiers for text-, image-, and audio-to-video generation. LTX 2.5 supports LoRA training and fine-tuning on custom datasets, making it adaptable to specific domains.
falAug 11fal launched the Mirage Avatar X API, providing access to Mirage’s most advanced avatar generation model. The API supports text-to-video and reference-to-video generation with identity preservation and natural expressivity. It is available now on the fal platform, with generation priced at $0.30 per second.
falAug 10fal launched a LoRA trainer for the MiniMax H3 video model, allowing custom adapter training on captioned video or image datasets. The trainer supports subject, object, and style learning with joint video and audio processing. fal also released an open-source Realism People LoRA to demonstrate the model's photorealistic human generation capabilities.
falAug 8fal launched fal Agent, a creative assistant that orchestrates generative media workflows across image, video, and 3D models. The agent selects models, maintains character and style consistency across generations, and manages multi-step tasks. It is available in early access via web, API, CLI, and MCP server, with Pro and Max subscription tiers starting at $200 per month.
falAug 8fal has launched ByteDance’s Seedance 2.5 model, enabling 30-second continuous video generation and multi-shot sequences in a single pass. The integration supports up to 50 multimodal reference inputs for character and style consistency, alongside improved prompt and 3D layout adherence. It is available now via text-to-video, image-to-video, and reference-to-video API endpoints.
falAug 8Fal now provides secure, U.S.-hosted access to ByteDance’s Dreamina Seedance and Seedream models through its enterprise platform. This expansion addresses demand from studios and enterprise customers for secure, domestic infrastructure for frontier AI models. The models are available to select U.S. organizations.
falAug 8fal integrated Math Magic’s Hi3D model family, enabling 3D asset generation from single images. The platform provides six endpoints, including image-to-3D and 3D texturing, supporting 1536³ resolution with high-precision geometry. These models produce production-ready assets with clean topology, now available for immediate inference through the fal API and playground.
falAug 5fal integrated Black Forest Labs’ FLUX.3 Keyframes-to-Video endpoint, which generates motion and synchronized audio between user-provided keyframe images. The service offers precise control over shot progression and is available for commercial use. Pricing is set at $0.17 per second for 720p resolution and $0.29 per second for 1080p output.
falAug 4fal has integrated Alibaba’s Qwen Image 3.0 Pro, a text-to-image model that handles dense typography, identity preservation, and complex image edits. The platform provides text-to-image and image-editing endpoints, priced at $0.04 per image at 1K resolution and $0.075 at 2K resolution.
falAug 3fal has integrated HeyGen’s Filler Word Removal API, enabling automated removal of filler words and dead air from video files. The service processes footage via a single API call, priced at $0.01 per second of input video with a 60-second minimum. Credits are not deducted if the model makes no cuts to the source material.
falAug 3fal has made the open weights for MiniMax H3 available for immediate inference on its platform. The integration supports text-to-video, image-to-video, and reference-to-video generation at 2K resolution. Each second of video output costs $0.26, with durations ranging from 5 to 15 seconds across seven aspect ratios.
falAug 1fal updated Grok Imagine Video 1.5 with native 1080p support for text-to-video and image-to-video generation. The platform also added a reference-to-video mode that supports up to seven reference images and an audio reference, maintaining character consistency at 720p. These updates expand the model's resolution and input control for generative video workflows.
falJul 31fal has integrated the MiniMax H3 multimodal model, enabling 2K video generation with native stereo audio. The platform provides text-to-video, image-to-video, and reference-to-video endpoints, with generation priced at $0.26 per second. The model supports complex creative tasks, such as composing synchronized soundtracks and animation from text prompts and still image inputs in a single generation pass.
falJul 30Fal launched its Product Photo API, powered by Pixelcut, to generate studio-quality images with 100% product fidelity. The API automates background removal, shadow generation, and logo placement for e-commerce catalogs. It processes images at $0.024 per request, providing a scalable solution for retailers and marketplaces needing consistent, professional product imagery without manual photoshoots.
falJul 24fal has integrated Alibaba Cloud’s Happy Oyster 1.0, a world model that generates interactive 3D environments from text. The new Realtime Playground supports adventure exploration and real-time directing with narrative consistency. World generation costs $2.00, while live sessions are priced at $0.014 per second with a 60-second minimum.
falJul 21fal has integrated Sonilo’s Sound Effects 1.0, a model that generates synchronized, royalty-free audio from video or text prompts. The platform provides three API endpoints for video-to-sound, video-to-video, and text-to-sound generation. The service is available for commercial use, priced at $0.009 per second of output.
falJul 17fal released a new workflow template that uses Video Depth Anything to extract consistent depth maps from source footage. This depth data guides ByteDance’s Seedance 2.0 to transform the video’s visual style while preserving original motion and composition. The workflow is available for use through the fal platform’s template library and API.
falJul 16fal has added the LTX-2.3 Reframe model to its platform, enabling video aspect ratio conversion without destructive cropping. The model recenters footage and generates content for newly exposed areas to match the original scene. It supports 16:9, 9:16, 1:1, 4:5, and 5:4 ratios for videos up to 60 seconds, with 720p and 1080p output resolutions.
falJul 16fal integrated Decart’s Lucy 2.5 Realtime, a model for live video-to-video editing over WebRTC. The API enables real-time restyling, background swapping, object manipulation, character swaps, and VFX application on webcam or streamed feeds. It features improved edit isolation for non-human objects and is available via the fal playground and API for $0.04 per second.
falJul 15fal has integrated the Bria AI Product Dimensions API, which generates marketplace-ready images with dimension callouts, labels, and weight or capacity readouts from a single product photo. The deterministic pipeline maintains pixel-accurate original photos without redrawing them. The service is available via API for $0.04 per image.
falJul 14fal has integrated VEED Studio’s Lip Sync 2.0, a model that automatically synchronizes lip movements, emotion, and style to new audio tracks. The API processes footage up to 4K resolution and clips up to 10 minutes without requiring training or fine-tuning. It features controllable sync strength and is available for $0.07 per second of output video.
falJul 13fal optimized Ideogram V4 inference to 0.44 seconds at 1K resolution, a 6x speedup. The pipeline uses FP4 kernel fusion, quantization-aware distillation, and timestep distillation to collapse guidance branches and denoising steps. Additionally, fal reached 1,000 tokens per second for prompt expansion using a custom DSpark speculative decoding model, keeping expansion under two seconds.
falJul 9fal achieved 0.45s inference for Ideogram V4 at 1K resolution, a 6x speedup over the baseline. The optimization pipeline uses FP4 kernel fusion to minimize memory round-trips, quantization-aware distillation to maintain color fidelity, and timestep distillation to reduce denoising steps. These techniques collapse classifier-free guidance into a single forward pass, delivering production-grade performance without quality loss.
falJul 8fal reached over 1,000 tokens per second for the Ideogram V4 prompt expander by training a custom DSpark adapter. This optimization delivers 16 times higher throughput and keeps prompt expansion under two seconds. The engineering effort is the first in a series detailing how the platform built its high-speed generative media infrastructure.
falJul 8fal has integrated ByteDance’s Seedream 5.0 Pro, a text-to-image model featuring region-precise editing and native text support in 14 languages. The model generates structured designs, including posters, product mockups, and UI layouts. It is available via API, priced at $0.0675 per image for resolutions up to 1536x1536 pixels and $0.135 for resolutions up to 2048x2048.
falJul 6fal has released two in-house speed-optimized variants of Ideogram V4, named Fast and Instant. The Instant model generates images in 0.4 seconds, while the Fast model takes one second, delivering up to 8x faster performance than the original. Both variants maintain the same prompt and parameter interface for immediate integration into existing workflows.
falJul 3fal integrated support for 3D, CG, and game-engine renders within its Seedance 2.0 reference-to-video API. The model converts these inputs into photorealistic 4K footage while preserving the original composition and camera movement. This workflow allows for the direct transformation of Blender or game-engine animations into high-fidelity video output.
falJun 30fal integrated Bria’s Extract Object model, which isolates specific objects from images using natural language prompts. The API returns clean RGBA-PNG cutouts with transparent backgrounds, eliminating the need for manual masking or editing. This model is available for commercial use on the fal platform, priced at $0.02 per image.
falJun 28fal released 3DREAL, an open-source IC-LoRA for the LTX-2.3 video model that transforms 3D, CG, and game-engine renders into photorealistic video. The adapter preserves the input's exact composition and camera movement. It ships in Light and Strong variants, available via Hugging Face weights or a ready-to-use serverless endpoint on fal’s platform.
falJun 25fal released seven new LTX 2.3 video effect LoRAs, including water simulation, day-to-night, and instant shave. These in-context adapters apply visual transformations to any video in a single pass. Each effect is available as a dedicated API endpoint on fal's platform, priced at $0.0024075 per megapixel of generated video data.
falJun 25fal has exclusively integrated ByteDance's Seed Audio 1.0, an all-in-one model that generates voice, music, and sound effects in a single pass. The model builds multi-speaker scenes from a single prompt and accepts up to three reference clips to guide voice and emotion. It is available for commercial use via API at $0.075 per minute of generated audio.
falJun 24fal has integrated ByteDance's Seedance 2.0 Mini, a video generation model that operates at twice the speed and half the cost of the standard version. The model maintains character, wardrobe, and visual style consistency across multi-shot sequences. It is available via text-to-video, image-to-video, and reference-to-video endpoints, with pricing starting at 0.0721 dollars per second for 480p output.
falJun 23fal has released the open-source Krea 2 image model, featuring open weights and code. The integration includes the Turbo variant for fast text-to-image generation at 0.008 dollars per megapixel and the RAW base checkpoint for custom LoRA training. These models are available via fal’s serverless API for inference and fine-tuning workflows.
falJun 22fal has integrated Sync Labs' sync-3 avatar model, which transforms a single still image into a talking character with lipsync synchronized to a provided voice track. The model supports any illustration or animated frame and is available via API and web playground for commercial use, priced at $8 per minute of generated video.
falJun 22fal has added native 4K resolution support for ByteDance's Seedance 2.0 model across its text-to-video, image-to-video, and reference-to-video endpoints. The update generates video with high-resolution detail and motion consistency, priced at 0.008 dollars per 1000 tokens. This expansion enables production-grade video generation directly through the fal serverless API.
falJun 22fal integrated Alibaba’s Happy Horse 1.1, a video model generating 1080p content with synchronized native audio and multilingual lip-sync. The platform supports text-to-video, image-to-video, and reference-to-video workflows using up to nine reference images. It is available via API and web playground, with pricing starting at 0.14 dollars per second for 720p output.
falJun 19fal launched the LTX 2.3 Ingredient endpoint, which generates video scenes with optional synced audio from a single reference sheet. The model processes characters, objects, props, and environments from one visual prompt, targeting creative ads and product demos. It is available via API at $0.0024075 per megapixel of generated video data.
Frequently asked questions