{ "meta": { "version": "1.0", "compiled": "2026-05-13", "note": "Service facts are compiled from public sources (vendor sites, Wikipedia, official announcements) and reflect best-available information at compile time. Verify against vendor docs before procurement." }, "services": [ { "id": "sora-2", "name": "Sora 2", "vendor": "OpenAI", "category": "frontier", "status": "in-transition", "released": "2025-09", "license": "proprietary", "specs": { "max_clip_seconds": 60, "max_resolution": "1080p", "native_audio": true, "image_to_video": true, "video_to_video": false, "character_consistency": "strong" }, "access": ["sora-app (sunsetting)", "chatgpt-pro", "api (planned wind-down)"], "pricing_tier": "premium", "strengths": [ "Best-in-class prompt adherence on cinematic scenes", "Native audio dialogue, ambience and SFX", "Cameo system for likeness-based generation", "Multi-shot continuity in a single prompt" ], "weaknesses": [ "Access has been turbulent — dedicated apps were sunset in 2026 and the standalone API is on a public wind-down path; treat as a moving target", "Watermarked output on most consumer tiers", "Hard content filters reject many commercial briefs", "Throughput is slow during peak hours" ], "best_for": "Cinematic prototyping, hero spots, mood pieces where prompt fidelity matters most.", "skip_if": "You need a stable, contractable API today, or your workflow centres on stock-style B-roll." }, { "id": "veo-3-1", "name": "Veo 3.1", "vendor": "Google DeepMind", "category": "frontier", "status": "generally-available", "released": "2025-05 (Veo 3); 3.1 update later in 2025", "license": "proprietary", "specs": { "max_clip_seconds": 8, "max_resolution": "4K", "native_audio": true, "image_to_video": true, "video_to_video": "via Flow", "character_consistency": "strong" }, "access": ["gemini", "flow", "google-ai-studio", "vertex-ai", "gemini-api"], "pricing_tier": "enterprise + consumer", "strengths": [ "Native audio (dialogue + SFX + ambience) is best-in-class", "4K output is a real differentiator for delivery", "Camera-control vocabulary (dolly, pan, zoom) actually works", "Vertex AI route makes enterprise procurement painless" ], "weaknesses": [ "8-second hard cap per clip — long-form means stitching", "Pricing concentrates in the $250/mo Ultra tier for unrestricted use", "Content policy on real-person likeness is strict", "Latency is variable depending on region and time of day" ], "best_for": "Brand films, product reveals, anything where on-clip audio matters.", "skip_if": "Your edit needs single takes longer than 8 seconds, or you cannot tolerate enterprise gating." }, { "id": "runway-gen-4-5", "name": "Gen-4.5 · Aleph", "vendor": "Runway", "category": "frontier", "status": "generally-available", "released": "2025 (Gen-4); Gen-4.5 follow-up", "license": "proprietary", "specs": { "max_clip_seconds": 10, "max_resolution": "1080p", "native_audio": "partial", "image_to_video": true, "video_to_video": true, "character_consistency": "best-in-class" }, "access": ["runway-web", "runway-api"], "pricing_tier": "studio + enterprise", "strengths": [ "Character and world consistency across shots is the category leader", "Aleph turns a video into a re-styled video — true video-to-video at quality", "Mature filmmaker-facing UI (timelines, references, motion brushes)", "Act-One / Act-Two performance capture is genuinely useful for character work" ], "weaknesses": [ "Credit-burn economics — costs scale fast on iteration", "Native audio still trails Veo and Sora", "1080p ceiling on the main path", "Output quality has high prompt-engineering ceiling" ], "best_for": "Episodic content, character-driven work, agencies doing reference-based generation.", "skip_if": "You need cheap-and-fast social cuts or 4K master delivery." }, { "id": "kling-2", "name": "Kling 2.x", "vendor": "Kuaishou", "category": "frontier", "status": "generally-available", "released": "2024-06 (1.0); 2.x family throughout 2025", "license": "proprietary", "specs": { "max_clip_seconds": 120, "max_resolution": "1080p", "native_audio": false, "image_to_video": true, "video_to_video": "limited", "character_consistency": "good" }, "access": ["klingai-web", "kwaicut-app", "api"], "pricing_tier": "consumer + studio", "strengths": [ "Up to two-minute generations at 30 fps 1080p — uncommon length", "Strong on physical realism, water, fabric, complex motion", "Image-to-video produces convincing results from a single still", "Aggressive pricing vs Western peers" ], "weaknesses": [ "Account creation historically required a Chinese phone number; English UX has improved but is still uneven", "No native audio generation", "Content policy is opaque and changes with little notice", "Long renders take real wall-clock time" ], "best_for": "Two-minute scenes, realistic motion, image-to-video at production length.", "skip_if": "You need audio baked in, or your compliance team will not sign off on a China-hosted pipeline." }, { "id": "pika", "name": "Pika · Pikaformance", "vendor": "Pika Labs", "category": "creator", "status": "generally-available", "released": "2023 (1.0); Pikaformance era 2025–2026", "license": "proprietary", "specs": { "max_clip_seconds": 10, "max_resolution": "1080p", "native_audio": "lip-sync + sfx", "image_to_video": true, "video_to_video": "via effects", "character_consistency": "good" }, "access": ["pika-web", "discord"], "pricing_tier": "consumer", "strengths": [ "Pikaffects and Scene Ingredients let non-technical users get a result fast", "Pikaformance lip-sync is among the most convincing in the consumer tier", "Near-real-time generation on shorter clips", "Lowest friction-to-first-good-clip of any tool here" ], "weaknesses": [ "Long-form is not its game — short clips only", "Prompt adherence is good not great vs frontier models", "Output resolution caps below Veo or open-source LTX-2", "Brand-safe rendering still requires curation" ], "best_for": "Social-first creators, meme-velocity content, talking-head shorts.", "skip_if": "Your brief is broadcast spots or anything over ~10 seconds without obvious cuts." }, { "id": "luma-ray", "name": "Ray 2 · Dream Machine · Agents", "vendor": "Luma", "category": "frontier", "status": "generally-available", "released": "2024 (Dream Machine); Ray 2 + Agents 2025", "license": "proprietary", "specs": { "max_clip_seconds": 10, "max_resolution": "1080p", "native_audio": true, "image_to_video": true, "video_to_video": true, "character_consistency": "good" }, "access": ["lumalabs-web", "api", "agents"], "pricing_tier": "consumer + studio + enterprise", "strengths": [ "Ray 2 motion is fluid and physically plausible", "Agents orchestrate image + video + audio + copy in a single brief", "Clear pricing ladder ($30 / $90 / $300) makes procurement simple", "Used in production by Publicis and other named agencies" ], "weaknesses": [ "10-second cap; long-form requires stitching or Agents orchestration", "API throughput tighter than Runway's at the same tier", "Output sometimes prioritises smoothness over prompt detail", "Ultra plan needed to unlock the full Agents stack" ], "best_for": "Agency teams running multi-modal campaigns, anyone who wants a single orchestrated workflow.", "skip_if": "You only need one-shot text-to-video and want the absolute lowest cost-per-clip." }, { "id": "hailuo", "name": "Hailuo", "vendor": "MiniMax", "category": "creator", "status": "generally-available", "released": "2024; rapid iteration through 2025–2026", "license": "proprietary", "specs": { "max_clip_seconds": 10, "max_resolution": "1080p", "native_audio": "partial", "image_to_video": true, "video_to_video": "limited", "character_consistency": "good" }, "access": ["hailuoai-web", "miniMax-api"], "pricing_tier": "consumer + studio", "strengths": [ "Strong physical realism on faces and bodies", "Bundled creative tools — PetPal, BabyForm, ASMR Generator, Style Switch — drop the prompt-craft barrier for normies", "Hailuo Agent automates multi-tool workflows", "Image-to-video quality holds up well at scale" ], "weaknesses": [ "English documentation lags Chinese-language docs", "Pricing on heavy use is not the clearest on the web", "Content policy enforcement is opaque", "Brand-safe enterprise contracting is harder than with Western vendors" ], "best_for": "Creator economy, social, persona videos and \"verticalised\" generators.", "skip_if": "You need a Western-vendor contract and SLA-grade support." }, { "id": "seedance-2", "name": "Seedance 2.0", "vendor": "ByteDance Seed", "category": "creator", "status": "generally-available", "released": "2026-02", "license": "proprietary", "specs": { "max_clip_seconds": 15, "max_resolution": "1080p", "native_audio": "partial", "image_to_video": true, "video_to_video": false, "character_consistency": "good" }, "access": ["bytedance-web", "api"], "pricing_tier": "consumer + studio", "strengths": [ "Strong dance / human-motion fidelity — the model has clearly been trained for it", "5–15 second range covers most short-form use cases", "Fast turnaround compared to frontier competitors", "TikTok-adjacent pipeline integration for creators already on ByteDance products" ], "weaknesses": [ "Public reporting in 2026 has flagged copyright-infringement concerns on training data — that may carry legal risk for commercial use", "Documentation is thin in English", "No native audio on the main path", "Enterprise procurement story is unclear outside China" ], "best_for": "Short-form social, motion-heavy content where dance and bodies are the subject.", "skip_if": "Your legal team is allergic to unresolved training-data provenance questions." }, { "id": "ltx-2", "name": "LTX-2", "vendor": "Lightricks", "category": "open-source", "status": "generally-available", "released": "2025-10", "license": "open-source", "specs": { "max_clip_seconds": 60, "max_resolution": "1080p", "native_audio": true, "image_to_video": true, "video_to_video": true, "character_consistency": "good" }, "access": ["weights (HF)", "ltx-studio", "self-host"], "pricing_tier": "free + studio", "strengths": [ "Open weights — full ownership of pipeline, no API rate limits", "60-second generations natively, not stitched", "Native audio built into the model rather than bolted on", "Strong VRAM efficiency for an open model — runs on prosumer GPUs" ], "weaknesses": [ "Self-hosting requires real GPU infrastructure to be production-grade", "Hosted LTX Studio costs less per clip than Runway but has fewer creative-control surfaces", "Output ceiling sits below Veo 3.1 / Sora 2 on hero shots", "Smaller plugin ecosystem than the proprietary tools" ], "best_for": "Teams that need ownership, on-prem deployment, or just want to escape per-clip pricing.", "skip_if": "You have no GPU budget and your team has zero appetite for self-hosting." }, { "id": "wan-2", "name": "Wan 2.x", "vendor": "Alibaba", "category": "open-source", "status": "generally-available", "released": "2025; ongoing 2.x releases", "license": "open-source", "specs": { "max_clip_seconds": 10, "max_resolution": "1080p", "native_audio": "partial", "image_to_video": true, "video_to_video": "limited", "character_consistency": "good" }, "access": ["weights (HF + ModelScope)", "alibaba-cloud", "self-host"], "pricing_tier": "free + cloud", "strengths": [ "Frontier-adjacent quality from an open model", "Multilingual prompt understanding — Mandarin briefs translate cleanly", "Image-to-video at quality that surprises people the first time", "Active community on ModelScope and HF" ], "weaknesses": [ "10-second cap on the main path", "Documentation skews to Chinese audiences", "Going from research checkpoint to production pipeline is real engineering work", "Native audio still partial" ], "best_for": "Self-hosted teams who want frontier-class quality without proprietary lock-in.", "skip_if": "You need a turnkey hosted product with a Western support contract." } ] }