Fal 在 2026年9月 发布的内容

ideogram/v4.5/edit, details and endpoints

Fal introduced Ideogram v4.5, enabling image edits with up to 4 reference images, optional masking, and a high-precision mode that preserves unchanged pixels. Pricing ranges from $0.008 to $0.22 per i…

ideogram/v4.5, details and endpoints

Fal.ai launched Ideogram v4.5, enhancing its image generation model with improved text rendering for posters and logos. The update introduces three pricing tiers: $0.03 per image for Low quality, $0.0…

pixelcut/looping-video, details and endpoints

Fal.ai released pixelcut/looping-video, a model that converts a single product image into a 5–15 second looping video with locked camera, ambient motion, or 360° spin. Pricing ranges from $0.08 to $0.…

veed/clean-audio, details and endpoints

Fal introduced veed/clean-audio, a model that converts noisy speech recordings into studio-quality audio. The service charges $0.0125 per input minute, rounded to the nearest minute with a 1-minute mi…

minimax/h3-max/insert-video, details and endpoints

Fal introduced a new H3 Max feature allowing users to insert AI-generated scenes into existing videos by specifying edit points and providing prompts, reference images, or videos. The feature bills at…

google/gemini-3.8-flash-tts, details and endpoints

Fal introduced Google’s Gemini 3.8 Flash TTS, enabling expressive speech generation with 30 voices, style instructions, and support for single-speaker narration or two-speaker dialogue. The model cost…

google/gemini-3.8-flash-lite-tts, details and endpoints

Fal.ai introduced Google’s Gemini 3.8 Flash Lite TTS model, enabling expressive speech generation with 30 voice options and style instructions for single-speaker narration or two-speaker dialogue. The…

bytedance/seedance-2.5/draft/complete, details and endpoints

Fal introduced a new draft completion endpoint for ByteDance’s Seedance 2.5 model, enabling users to regenerate video tasks at 1080p resolution by submitting a draft ID. The endpoint costs $0.0214 per…

fal-ai/flux-3-action/so101

Fal introduced the Flux-3-Action SO101 model, a new endpoint designed for robotics control tasks. The model charges $0.003 per compute second, with typical calls costing $0.0015–$0.006 per invocation,…

minimax/h3-max/extend-video

Fal.ai’s minimax/h3-max/extend-video model now supports extended video generation with tiered pricing: $0.05/s at 480p, $0.08/s at 768p, $0.16/s at 1080p, and $0.32/s at 2K. Each request includes 4,09…

minimax/h3-max/3d-to-video

Fal introduced a new 3D-to-video generation model, minimax/h3-max/3d-to-video, enabling users to convert 3D inputs into video. The model charges $0.75–$1.30 for the first 5 seconds (480p–1080p) and $0…

recraft/v4.1/flash/text-to-image

Fal.ai introduced recraft/v4.1/flash, a new text-to-image model generating high-quality raster images from prompts with controls for size, color, and background. The model supports commercial use and …

bytedance/seedream/v5/flash/edit

Fal.ai introduced bytedance/seedream/v5/flash/edit, a new image editing model optimized for speed and cost efficiency. The model supports commercial use and charges $0.027 per image, targeting workflo…

fal-ai/elevenlabs/tts/multilingual-v2

Fal now offers ElevenLabs TTS Multilingual v2, enabling multilingual text-to-speech audio generation through its platform. The model supports commercial use, streaming, and API integration, with a cos…

fal-ai/ace-step/prompt-to-audio

Fal introduced ACE-Step, a new model enabling text-to-audio generation from simple prompts, such as creating lofi hiphop music. The feature is available via API and playground with commercial use perm…

fal-ai/meshy/v6-preview/image-to-3d

Fal introduced Meshy-6-Preview, a new AI model generating production-ready 3D assets from images. The feature is available via Fal’s API for $0.8 per generation, supporting commercial use and multiple…

fal-ai/kling-video/v2.5-turbo/standard/image-to-video

Fal introduced Kling Video v2.5-turbo, an upgraded image-to-video model with enhanced input flexibility (drag-and-drop, URLs, clipboard) and broader file type support. The new model charges $0.21 for …

fal-ai/kling-video/v3/turbo/pro/image-to-video

Fal introduced the Kling Video v3 Turbo Pro image-to-video model, enabling users to generate video from images. The service charges $0.14 per second of video output, with a 5-second video costing $0.7…

tripo3d/p2/text-to-3d

Fal introduced explicit pricing tiers for its tripo3d/p2 text-to-3D model, ranging from $1.00 per model without textures to $1.30 with extreme textures. PBR textures and quad topology are available at…

tripo3d/p2/image-to-3d

Fal.ai launched tripo3d/p2, an image-to-3D model generation feature with transparent per-model pricing. Costs range from $1.00 per model without textures to $1.30 with extreme textures, with PBR and t…

pixverse/music-video/vibemv

Fal introduced VibeMV, an AI model generating music videos from audio inputs, priced at $0.06/second for 720p and $0.09/second for 1080p. The tool accepts MP3, OGG, WAV, M4A, or AAC files and enforces…

minimax/h3-max/styles/retro-toon-70s

Fal introduced a new 'Retro Toon 70s' style for its H3 Max Director model, generating 768p videos with audio in a hand-painted 1970s animation aesthetic. The style supports 5–15 second durations and c…

minimax/h3-max/styles/low-poly

Fal introduced a new low-poly 3D style for its H3 Max video generation model, producing 768p videos with audio from text prompts or images. Pricing is $0.08 per second, with 5-second videos costing $0…

minimax/h3-max/styles/hand-drawn

Fal.ai expanded its H3 Max video generation model with a new 'hand-drawn' animation style, producing 768p videos with audio from text prompts or first-frame images. Pricing is $0.08 per second, with 5…

minimax/h3-max/styles/16bit-pixel

Fal introduced the H3 Max Director model, enabling 768p video generation with audio in a 16-bit pixel-art style from text prompts or first-frame images. Videos range from 5–15 seconds, priced at $0.08…

meshy/v7.1/multi-image-to-3d

Fal’s meshy v7.1 model now supports multi-image-to-3D generation with tiered pricing: $0.80 for a base model, $1.20 with textures, $1.52 for textured models with auto-rigging and animation, and $1.40 …

meshy/v7.1/image-to-3d

Fal released meshy v7.1 for image-to-3D conversion, introducing tiered pricing: $0.80 for base models, $1.20 with textures, $1.52 with textures plus auto-rigging and animation. Users can now customize…

meshy/v7.1/text-to-3d

Fal.ai’s Meshy v7.1 text-to-3D model now supports untextured and textured outputs, Smart Topology, and geometry resolutions up to 4K. New pricing tiers include $0.80 for base models, $1.20 for texture…

minimax/h3-max/styles/vhs

Fal.ai introduced a new model, minimax/h3-max/styles/vhs, which generates 768p VHS-style videos with audio from text prompts or an image. The model supports 5–15 second clips and allows adjustable tap…

google/lyria-3.5

Fal.ai now hosts Google DeepMind’s Lyria 3.5, enabling users to generate commercial-ready music via text prompts. The model supports detailed inputs like genre, tempo, and instrumentation, with pricin…

bytedance/seedance-2.5/us/image-to-video

Fal introduced ByteDance’s seedance-2.5 image-to-video model, enabling users to convert images into videos with customizable controls. Pricing is $0.5676/second for 720p and $0.2646/second for 480p, w…

bytedance/seedance-2.5/us/text-to-video

Fal.ai now charges $0.5676 per second for 720p video and $0.2646 per second for 480p video generated via the bytedance/seedance-2.5/us/text-to-video model. Token costs are $0.02568 per 1000 tokens for…

bytedance/seedance-2.5/us/reference-to-video

Fal introduced the ByTEDance Seedance 2.5 model for video generation with a reference-to-video input feature, allowing users to generate videos from images, videos, or audio inputs. Pricing is structu…

minimax/h3-max/lip-sync/image-to-video

Fal introduced a new AI model, minimax/h3-max/lip-sync/image-to-video, enabling users to generate lip-synced videos from static images and audio inputs. The model supports multiple file types and offe…

H3 Max: Built with fal Inference and Training

fal.ai introduced H3 Max, a video generation model that achieves 5-second video generation in under 3 seconds while ranking first in quality and prompt understanding across benchmarks. The model lever…

fal-ai/elevenlabs/tts/eleven-v3

Fal.ai now offers ElevenLabs' Eleven v3 text-to-speech model on its platform, enabling lifelike, emotionally nuanced audio generation with inline audio tag control and support for 70+ languages. The i…

bria/fibo-edit-1.5/virtual-try-on

Fal introduced a new virtual try-on model, bria/fibo-edit-1.5/virtual-try-on, enabling users to generate try-on images from input photos. The model charges a flat $0.04 per generated image, with bulk …

async/tts-pro/v1.0

Fal introduced async/tts-pro/v1.0, a new text-to-speech model enabling professional-quality voiceovers with granular control over pauses, emphasis, and timing. The model supports commercial use and ch…

bytedance/seedance-2.0/us/reference-to-video

Fal introduced ByteDance’s seedance-2.0 model for reference-to-video generation, supporting images, videos, and audio inputs. Pricing is $0.37/second for 720p and $0.1731/second for 480p, with token-b…

bytedance/seedance-2.0/us/image-to-video

Fal introduced ByteDance’s seedance-2.0 image-to-video model, enabling users to generate videos from images with customizable inputs. Pricing is $0.37/second for 720p and $0.1731/second for 480p, with…

blackforestlabs/flux-3/image-to-video

Fal introduced a new image-to-video API powered by Black Forest Labs’ Flux 3 model, enabling video generation from a single input image and text prompt. The API supports configurable aspect ratios, re…

fal-ai/kling-video/v3/pro/image-to-video

Fal introduced Kling V3 (pro) for its image-to-video API, improving visual quality, motion consistency, and adding native audio generation in Chinese and English. The update includes multi-shot video …

fal-ai/krea-2/turbo

Fal.ai deprecated the @fal-ai/serverless-client package in favor of @fal-ai/client for its krea-2/turbo API, requiring users to update imports and set FAL_KEY for authentication. The API now supports …

bytedance/seedance-2.0/us/text-to-video

Fal now lists pricing for ByteDance’s Seedance 2.0 text-to-video model on its platform, charging $0.37/second for 720p and $0.1731/second for 480p video generation. Token costs are $0.0168 per 1000 to…

creatify/boreal

Fal introduced creatify/boreal, an AI model that generates synchronized native-audio videos from text with optional image and audio inputs. The model supports commercial use and offers tiered pricing:…

elevenlabs/music/v2

Fal introduced elevenlabs/music v2, offering users more control over input customization for music generation. The model now charges $0.6 per output audio minute, rounded up to the nearest minute, rep…

elevenlabs/music/v2.5

Fal released elevenlabs/music v2.5, introducing enhanced input customization controls for music generation. The update also introduces a new pricing model of $0.6 per output audio minute, rounded up t…

fal-ai/meshy/v6-preview/text-to-3d

fal.ai introduced Meshy v6-preview, a text-to-3D generation model with enhanced customization controls. The new version costs $0.80 per generation, replacing or updating prior iterations of the model.…

meshy/v7/multi-image-to-3d

Fal.ai released meshy v7, enabling users to generate textured 3D models from multiple input images for $1.20 per call, or $1.40 with ultra mode. Optional auto-rigging and animation features add $0.20 …

minimax/h3-max/camera-controls

Fal.ai introduced camera controls for its minimax/h3-max video generation model, allowing users to drag the camera, adjust distance via scroll, and look around by dragging the background. The update i…

tripo3d/h3.1/multiview-to-3d

Fal.ai introduced tripo3d/h3.1, a new model enabling conversion of multiview images into 3D models. Users can generate models at three pricing tiers ($0.20–$0.40) with optional add-ons for textures, d…

blackforestlabs/flux-3/edit-video

Black Forest Labs introduced Flux-3, a new AI-powered video editing model enabling users to edit videos via text prompts or direct file uploads. The model charges $0.03 per second for 720p output and …

fal-ai/hunyuan-3d/v3.1/rapid/text-to-3d

fal.ai released Hunyuan-3D v3.1, a rapid text-to-3D AI model with customizable input controls. The base generation costs $0.225 per use, while enabling PBR materials adds $0.15 per generation. Users c…

fal-ai/hyper3d/rodin/v2.5/text-to-3d

fal.ai released Hyper3D Rodin v2.5, a text-to-3D generation model with enhanced customization controls. The base generation costs $0.4 per request, while the HighPack addon increases the price to $1.2…

minimax/h3-max/multi-angle/image-to-video

Fal.ai introduced the minimax/h3-max/multi-angle image-to-video model, enabling users to generate multi-angle videos from static images with adjustable camera controls. The model supports multiple inp…

imagineart/imagineart-2.0-preview/text-to-image

Fal.ai introduced imagineart 2.0, a major upgrade to its text-to-image model, now available in preview. The new version offers enhanced customization and control for input prompts. Pricing is set at $…

fal-ai/ideogram/remove-background

fal.ai introduced Ideogram, a new AI-powered background removal model offering high-precision results with clean edges and transparent PNG output. The tool is accessible via fal.ai's serverless API or…

bria/video/background-removal/v3

Fal.ai released bria/video/background-removal/v3, a major upgrade to its video background removal model, now available for processing videos via file upload, URL, or clipboard paste. The new version i…

fal-ai/imageutils/rembg

Fal.ai introduced Imageutils Remove Bg, an AI-driven tool to automatically remove backgrounds from images with clean cutouts. The model supports common formats (JPG, PNG, etc.) and is available at $0 …

fal-ai/bria/background/remove

fal.ai introduced Bria RMBG 2.0, a production-grade background removal model priced at $0.018 per inference, trained exclusively on licensed commercial data to ensure compliance for enterprise use. Th…

fal-ai/birefnet

fal.ai introduced BiRefNet v2, a bilateral reference framework for high-resolution image segmentation with dual-reference architecture, improving edge detection and foreground-background separation. T…

fal-ai/veo3.1/fast/first-last-frame-to-video

fal.ai introduced veo3.1/fast, a new model enabling video generation from first and last frames. Pricing starts at $0.10/second (720p/1080p without audio) and scales to $0.35/second (4K with audio).…

fal-ai/recraft/v4/pro/text-to-image

Fal AI introduced recraft/v4/pro, a new text-to-image model with enhanced input customization controls. The model is priced at $0.25 per generated image, positioning it as a premium offering in the ge…

openai/gpt-image-2.5/sunburst/edit

Fal.ai published pricing for OpenAI’s gpt-image-2.5/sunburst/edit model, introducing tiered token-based costs: $5.00 per 1M input text tokens, $8.00 per 1M input image tokens, and $10.00–$30.00 per 1M…

openai/gpt-image-2.5/sunburst/text-to-image

Fal.ai introduced pricing for OpenAI’s GPT Image 2.5 text-to-image model, with costs varying by input/output tokens and quality settings. Input text tokens cost $5.00 per 1M, cached text $1.25, and ou…

openai/gpt-image-2.5/flare/edit

Fal.ai published pricing for OpenAI’s gpt-image-2.5 model, breaking down costs by text and image tokens with input/output tiers. Longer prompts, complex requests, and larger images increase costs, wit…

openai/gpt-image-2.5/flare/text-to-image

Fal.ai introduced new pricing for its OpenAI-compatible text-to-image model GPT-Image-2.5, with text tokens priced at $5.00 per 1M input, $1.25 cached, and $10.00 output, while image tokens cost $8.00…

bria/ad-delayer

Fal.ai introduced bria/ad-delayer, a new AI model enabling users to process and edit ad images with customizable layers. The tool supports drag-and-drop image uploads and costs $0.30 per request.

fal-ai/wan-25-preview/text-to-video

fal.ai introduced WAN-25-preview, a text-to-video model offering 480p, 720p, and 1080p outputs at $0.05, $0.10, and $0.15 per second respectively. The model accepts audio inputs via file upload, URL, …

MiniMax H3 Max Text to Video

fal.ai introduced H3 Max and H3 Max Turbo, post-trained variants of MiniMax H3, optimized for stronger prompt adherence, better aesthetics, and higher throughput via a custom inference stack without c…

minimax/h3-max/text-to-video

Fal.ai introduced the minimax/h3-max text-to-video model, offering video generation at $0.0125/second (480p) and $0.02/second (768p) during a limited-time promotional launch. Standard rates will rise …

Black Forest Labs introduced an image-to-video feature for its Flux-3 model, enabling users to convert static images into videos with customizable controls. The feature is priced at $0.17/second for 7

bytedance/seedance-2.5/image-to-video

ByteDance released Seedance 2.5, a next-generation AI image-to-video model on Fal.ai, enabling 30-second cinematic videos from a single image with native audio, physics, and camera control. The model …

minimax/h3/image-to-video

Fal AI introduced minimax/h3, a new image-to-video model that converts static images into videos. The model supports multiple input formats and offers four resolution tiers priced from $0.05 to $0.16 …

minimax/h3-max/image-to-video

Fal.ai introduced minimax/h3-max, a new image-to-video model, enabling users to convert images into videos at 480p ($0.0125/sec) or 768p ($0.02/sec) during a limited-time promotional launch. Standard …

fal-ai/nano-banana-2/edit

fal.ai introduced Nano Banana 2 Edit, an image-to-image model built on Google’s Gemini 3.1 Flash Image architecture, enabling semantic edits without manual masks. It supports up to 14 reference images…