VEED Subtitles API (veed/subtitles) is an AI-powered video subtitle generation and styling model that automatically transforms raw videos into polished, publish-ready content with professionally rendered burned-in subtitles. Designed for creators, marketers, and media teams, it combines automatic transcription, subtitle synchronization, and cinematic visual styling into a single streamlined workflow.

HiDream O1 Image Dev Edit (fal-ai/hidream-o1-image/dev/edit) is a lightweight, high-speed AI image editing and personalization model optimized for rapid reference-guided image transformations, creative experimentation, and cost-efficient editing workflows. Built on the HiDream O1 unified architecture, the dev edit variant enables users to edit, personalize, and transform images using prompts and reference images while maintaining strong subject consistency and high-resolution output quality.

HiDream O1 Image Edit (fal-ai/hidream-o1-image/edit) is a high-resolution AI image editing and personalization model built for advanced reference-guided image transformation, subject preservation, and commercial-quality creative editing workflows. Powered by the unified HiDream O1 architecture, the model performs intelligent image modifications using one or more reference images while maintaining strong visual consistency, identity fidelity, and prompt adherence. Unlike traditional image editing pipelines that rely on separate inpainting, identity adapters, or external personalization modules, HiDream O1 Image Edit handles:

HiDream O1 Image (fal-ai/hidream-o1-image) is a unified, production-grade AI image generation and editing model built for high-quality text-to-image creation, image editing, and personalized subject generation within a single native architecture. Supporting resolutions up to 2K, the model combines strong prompt understanding, subject consistency, commercial-grade realism, and flexible reference-guided workflows without relying on external adapters or identity modules. Unlike specialized pipelines that separate generation, editing, and personalization into different systems, HiDream O1 handles all major creative image tasks natively, making it highly efficient for scalable AI creative platforms, commercial design pipelines, and professional visual production workflows.

HiDream O1 Image (fal-ai/hidream-o1-image/dev) is a unified, multi-purpose AI image generation and editing model capable of handling text-to-image generation, image editing, and subject-driven personalization within a single native architecture. Designed for flexible creative workflows, it generates high-resolution images up to 2K resolution while maintaining strong visual consistency, prompt accuracy, and personalized subject fidelity. Unlike specialized pipelines that require separate models for generation, editing, or identity preservation, HiDream O1 combines all workflows into one streamlined model, making it highly effective for dynamic content creation systems and AI creative platforms.

Recraft V4.1 Utility Text-to-Image (fal-ai/recraft/v4.1/utility/text-to-image) is a lightweight, high-speed text-to-image generation model optimized for rapid creative workflows, large-scale ideation, and cost-efficient content production. Built on Recraft’s design-focused architecture, it maintains strong composition quality and aesthetic consistency while prioritizing faster generation throughput and scalable asset creation.

Recraft V4.1 Utility Pro Text-to-Image (fal-ai/recraft/v4.1/utility/pro/text-to-image) is a high-efficiency text-to-image generation model that combines the premium visual quality of Recraft V4.1 Pro with a more optimized, cost-effective runtime for scalable creative production workflows. It is specifically designed for teams and studios generating large volumes of professional-grade raster images without sacrificing visual polish or design consistency.

Recraft V4.1 Text-to-Vector (fal-ai/recraft/v4.1/text-to-vector) is a professional AI SVG vector generation model that converts text prompts into fully editable vector artwork with clean geometry, structured layers, and scalable design precision. Optimized for modern design workflows, it produces production-ready SVG graphics that can be directly edited in tools like Figma, Adobe Illustrator, Sketch, and CorelDRAW.

Recraft V4.1 Text-to-Image (fal-ai/recraft/v4.1/text-to-image) is a design-focused AI image generation model optimized for creating clean, production-ready raster images with strong prompt accuracy, balanced composition, and professional visual aesthetics. Built on Recraft’s design-first architecture, it excels at generating polished visuals suitable for branding, editorial content, marketing creatives, and modern digital design workflows.

Recraft V4.1 Pro Text-to-Vector (fal-ai/recraft/v4.1/pro/text-to-vector) is a premium AI vector generation model designed to create fully editable, high-quality SVG illustrations and scalable vector graphics directly from text prompts. Built for professional design workflows, it produces structurally clean vector compositions optimized for branding, print, UI assets, posters, icons, and commercial illustration work. Unlike raster image generators, this model generates resolution-independent SVG outputs that preserve geometric precision, clean paths, and scalable detail without quality loss. It is specifically optimized for producing visually balanced vector artwork suitable for editing in tools like Adobe Illustrator, Figma, CorelDRAW, and other vector-based design software.

Restore old or damaged photos by fixing colors, scratches, and resolution.

Change hairstyles and hair colors in photos realistically.

Balanced open-source text-to-image foundation model with 17B parameters using 28 inference steps, optimized for photorealistic photography and professional portrait generation. Excels at producing stunning photographic-quality images that set it ahead of competitors in realistic styles. Features unique four-encoder system (CLIP-L, CLIP-G, T5-XXL, Llama-3.1-8B) allowing specialized prompting strategies - CLIP for tag lists, T5/Llama for natural language descriptions. Standout capability for prompt evolution - small prompt changes don't reset the entire image, making it easier to iteratively refine compositions without starting over. Superior at text rendering within images for signs, labels, and branded content. Achieves excellent color restoration, edge processing, and detailed texture rendering. Particularly strong at realistic photographic styles, studio portraits, and professional photography aesthetics. Ideal for photographers, portrait artists, professional content creators, realistic artwork, fashion photography, product photography, studio-quality images, and creative projects requiring photorealistic outputs with iterative refinement workflows. Perfect balance between quality and generation time for professional applications. Released under MIT license for personal, research, and commercial use.

Vidu Reference-to-Image creates images by using a reference images and combining them with a prompt.

Ultra-fast 6B parameter text-to-image generator with LoRA support, developed by Tongyi-MAI. Z-Image Turbo LoRA delivers hyper-realistic portraits and detailed imagery in just 8 inference steps with optional acceleration modes (up to 40% faster). Excels at close-up portraits, ethnic/cultural photography, intricate skin textures, film grain aesthetics, and documentary-style imagery. Supports up to 3 custom LoRA weights for style adaptation, batch generation (1-4 images), and optional prompt expansion. Specialized for rapid prototyping, budget-conscious projects, detailed portrait photography, cultural/tribal imagery, and film camera aesthetics. Best for: fast generation, turbo speed, quick image, rapid prototyping, hyper-realistic portrait, close-up portrait, skin texture, pores, wrinkles, tribal, cultural photography, ethnic portrait, film grain, Kodak Portra, Leica, documentary, batch generation, low-cost, efficient, and LoRA styling. Category: Ultra-fast text-to-image with LoRA support and portrait specialization.

Bagel is a 7B parameter from Bytedance-Seed multimodal model that can generate both text and images.

This model reframes image editing as a video generation task using temporal reasoning tokens to ensure physically plausible transformations, maintaining consistency in lighting, shadows, reflections, and spatial relationships by imagining intermediate frames between original and edited states. Perfect for autonomous vehicle simulation, robotics training data, product visualization requiring physical accuracy, and any editing scenario where maintaining realistic physics and temporal coherence is critical.

Kling 2.1 Standard is a cost-efficient for the Kling 2.1 model, delivering high-quality image-to-video generation

Seedance 1.0 Lite

Advanced AI video generation model by Luma Labs specialized in creating photorealistic, action-packed content with natural coherent motion and cinematic quality. Excels at wildlife scenes featuring wild horses galloping across dusty desert plains, manes flying in wind, blazing midday sun, epic wide tracking shots with dynamic motion, and warm natural lighting - capturing lifelike animal movements and environmental interactions. Outstanding at high-speed action sequences including car chases, dynamic human movements, athletic performances, and fast-paced activities with ultra-realistic details and smooth motion fidelity. Features advanced cinematography controls supporting sweeping panoramas, intimate close-ups, dynamic tracking shots, crane movements, and precise camera dynamics specified through natural language (crane down, tracking shot, camera circles). Revolutionary keyframes feature enables frame-by-frame control with precise start and end frame specification for smooth narrative transitions and custom storytelling. Extend capability stretches videos beyond initial duration for building immersive narratives and complex sequences. Loop creation feature seamlessly blends video end with beginning for perfect looping content. Generates videos with lifelike textures, realistic lighting, physically accurate object interactions, fast coherent motion, logical event sequences, and substantially production-ready quality. Supports multiple aspect ratios with flexible loop options. Ideal for film previews, animation prototyping, action sequences, wildlife documentaries, marketing campaigns, promotional content, social media videos, VFX pre-visualizations, realistic backgrounds, cinematic b-roll, visual effects sequences, storytelling projects, and professional video production. Perfect for filmmakers, content creators, VFX artists, advertising agencies, wildlife videographers, action content producers, and creative professionals requiring photorealistic video generation with advanced camera control, natural motion physics, and cinematic storytelling capabilities for production-ready outputs.

Transform your images into stylized videos using this workflow.

Generate video clips from your images using Kling 1.6 (pro)

Framepack is an efficient Image-to-video model that autoregressively generates videos.

Framepack is an efficient Image-to-video model that autoregressively generates videos.

Lux TTS is a high-quality voice cloning text-to-speech model that converts text into natural-sounding speech using a reference audio sample to replicate the voice. The model generates 48kHz studio-quality audio while preserving the tone, pitch, speaking style, and vocal identity from the reference voice. The system is optimized for fast inference using a distilled architecture with only 4 inference steps, enabling efficient speech generation with minimal latency while maintaining strong voice fidelity. By analyzing a short reference audio clip, the model learns the voice characteristics and applies them when synthesizing speech from the provided text prompt. Lux TTS is ideal for AI voice cloning applications, narration generation, content creation, virtual assistants, and automated voice production pipelines where a consistent and realistic voice output is required. The model supports adjustable parameters for inference steps, voice adherence strength, and reference audio length, allowing developers to balance speed, voice accuracy, and generation diversity.

Google's latest advanced music generation model capable of creating any type of music from text descriptions. Specializes in producing high-quality instrumental tracks, ambient soundscapes, and professional background music across all genres including electronic, classical, jazz, rock, orchestral, and more. Perfect for video content creators, game developers, podcasters, and app designers needing royalty-free original music. Generates complete musical compositions with natural instrumentation, realistic melodies, harmonies, and atmospheric elements like nature sounds and ambient textures. Supports negative prompts for precise control over unwanted elements (vocals, tempo, style). Ideal for YouTube videos, social media content, game soundtracks, meditation music, podcast intros/outros, advertising, presentations, and any project requiring custom background music. Deterministic generation with seed control ensures reproducible results. Creates professional-quality tracks without requiring musical expertise or expensive licensing fees.

MiniMax Speech-02 Turbo - STABLE LEGACY high-speed text-to-speech with PROVEN RELIABILITY and up to 5000 CHARACTERS support. Delivers FAST GENERATION with consistent quality across 35+ LANGUAGES including Chinese, Japanese, Korean, Arabic, Hindi, and European languages. Features SPEED, VOLUME, and PITCH controls with PRONUNCIATION DICTIONARY for custom terminology and LANGUAGE BOOST for enhanced recognition. Battle-tested model ideal for PRODUCTION ENVIRONMENTS, established workflows, legacy integrations, and users preferring PROVEN STABILITY over newest features. Best choice when user needs RELIABLE FAST TTS with long text support and doesn't require latest features. Ultra-fast generation (3-15s).

MiniMax Speech 2.5 HD - HIGH-DEFINITION text-to-speech optimized for LONG-FORM CONTENT with up to 5000 CHARACTERS per request. Supports 40+ LANGUAGES including Persian, Filipino, Tamil, Chinese, Japanese, Korean, Arabic, Hindi, and European languages with native pronunciation boost. Delivers STUDIO-QUALITY audio with full SPEED, VOLUME, and PITCH controls plus PRONUNCIATION DICTIONARY for specialized terminology. Features ENGLISH NORMALIZATION for consistent pronunciation. Perfect for LONG ARTICLES, DOCUMENTS, EBOOKS, BLOG POSTS, and extended narration requiring HD audio quality. Best choice when user needs to convert LARGE TEXT BLOCKS or long-form content to premium speech. Medium generation (10-30s).

MiniMax Music v1 - REFERENCE-BASED song generator that creates music MATCHING THE STYLE of an uploaded audio sample. Requires a REFERENCE AUDIO FILE (15+ seconds, .wav/.mp3) containing music and vocals to establish the musical style, genre, and mood. Input your LYRICS (max 600 chars) and the AI generates a NEW SONG in the SAME STYLE as your reference. Use ## markers around sections for accompaniment/instrumental parts and newlines for pauses. Perfect for COVER-STYLE songs, creating music in a specific artist's style, matching existing brand music, style-consistent jingles, and generating songs that sound like a reference track. Best choice when user UPLOADS A SONG and wants NEW LYRICS sung in that style. Slow generation (60-120s).

ElevenLabs TTS Turbo v2.5 - HIGH-SPEED LOW-LATENCY text-to-speech optimized for REAL-TIME applications. Generates speech significantly FASTER than standard models while maintaining ElevenLabs quality - perfect for conversational AI, live interactions, chatbots, virtual assistants, and any application where RESPONSE TIME is critical. Same 20+ premium voices (Rachel, Aria, Roger, etc.) with stability, style, and speed controls. Supports multilingual via language codes and CONTINUITY CONTROLS for seamless audio concatenation. Word-level timestamps available for lip-sync. Best choice when user needs FAST TTS, REAL-TIME voice, or LOW-LATENCY speech generation. Instant generation (1-5s).

Lyria 3 Pro is an advanced AI music generation model designed to create high-quality original music directly from text prompts, supporting instrumentals, vocals, lyrics generation, multilingual singing, and image-inspired music composition. Built for creators, developers, and production workflows, it transforms descriptive prompts into fully produced audio tracks with strong