🟢 Seedance 2.0 NEW

Priority Tool Background Replacement

• Release: April 2026 — ByteDance's flagship AI video generation model
• Key Features:
  • Background replacement preserving character identity, camera movement, and performance
  • Support for image + text prompt to video (5s and 10s clips)
  • Character consistency across shots via reference image input
  • Camera motion control (pan, tilt, dolly, orbit)
  • Native audio generation for dialogue and sound effects
  • Anti-slop motion refinement for natural movement
• Access: Available via Doubao app (China), international access through API/partners
• VFX Relevance: Best-in-class for background replacement work — your primary use case

🟢 Kling AI UPDATED

Priority Tool Master Model

• Current Version: Kling 1.6 / Master model (April 2026)
• Key Updates:
  • Master model: Higher quality generation with better motion coherence
  • Improved character consistency across multi-shot sequences
  • Lip sync feature for dialogue scenes
  • Motion brush for directed character movement
  • Extended duration support (up to 10s at 1080p)
• API: Available via Kling Open Platform — REST API with SDK
• Pricing: Credit-based; Pro tier ~$7/mo for 660 credits, Premier ~$23/mo
• VFX Relevance: Strong competitor to Seedance for video generation; lip sync is unique

🟢 Runway Gen-4 UPDATED

Priority Tool Character Consistency

• Current Version: Gen-4 (released April 2026)
• Key Features:
  • Reference image system for consistent character/environment across shots
  • Scene-level control: define location + character + action separately
  • Camera direction with natural language prompts
  • 10s generation at 720p/1080p
  • Inpainting/outpainting for video regions
• API: Available — REST API for enterprise, with SDKs
• Pricing: Standard $15/mo (125 credits), Pro $35/mo (500 credits), Unlimited $95/mo
• VFX Relevance: Best reference system for multi-shot consistency; inpainting useful for cleanup

🔵 OpenAI Sora LAUNCHED

• Status: Available in ChatGPT Plus/Pro since December 2024, continued updates through April 2026
• Key Features:
  • Text-to-video up to 20s, image-to-video, video-to-video remix
  • Storyboard mode for multi-scene generation
  • Loop and blend features for seamless transitions
  • Max resolution 1080p
• Pricing: Included in ChatGPT Plus ($20/mo, 50 vids/mo), Pro ($200/mo, unlimited + higher res)
• Limitations: No API yet; watermark on free/Plus tier; physics sometimes off
• VFX Relevance: Good for concept/previs; not production-ready for VFX pipelines yet

🔵 Pika UPDATED

• Current Version: Pika 2.0+ (April 2026)
• Key Features:
  • Pika Effects: Incredibly diffuse, melt, explode, crush, and more visual effects
  • Lip sync with uploaded audio
  • Scene edit: Modify specific regions while preserving the rest
  • Outpainting to extend video frames
• Pricing: Free tier (250 credits), Standard $8/mo, Pro $28/mo, Unlimited $70/mo
• VFX Relevance: Pika Effects are unique for VFX-style transformations; scene edit useful for comp work

🔵 Luma Dream Machine UPDATED

• Current Model: Ray2 (April 2026)
• Key Features:
  • Ray2: Improved motion quality and physical plausibility
  • Camera motion control (orbit, pan, dolly)
  • Keyframe animation for precise motion control
  • LoRA training for style/character consistency
• API: Available via REST API
• Pricing: Free tier (30 gens/mo), Standard $24/mo, Pro $76/mo
• VFX Relevance: Camera control is excellent; LoRA training for consistent styles

🟡 Hailuo AI (Minimax) UPDATED

• Current Version: Hailuo/Minimax Video-01 (April 2026)
• High-quality text-to-video with strong motion coherence
• Subject reference feature for character consistency
• Available via API through Minimax platform
• Competitive quality with Kling at lower price points

🟡 Vidu UPDATED

• Current Version: Vidu 1.5+ (2026)
• Character reference for consistent subjects
• Fast generation speed (4s clip in ~30s)
• Available via API
• Strong in Asian market; growing international presence

🟡 New Entrants to Watch

• Haiper 2.0: Improved text-to-video with better temporal consistency
• Google Veo: Google's video generation model, available through Google AI Studio/Vertex
• Stable Video Diffusion 2.0: Open-source video generation, improving rapidly

OpenAI — GPT-4.1 NEW

• Released: April 14, 2026
• Three tiers: GPT-4.1, GPT-4.1 mini, GPT-4.1 nano
• All have 1M token context window
• Better instruction following & coding vs GPT-4o, cheaper too
• Pricing: $2.00/$8.00 (flagship), $0.40/$1.60 (mini), $0.10/$0.40 (nano) per 1M in/out
• GPT-5 still in preview — no public release date

Anthropic — Claude Opus 4 & Sonnet 4 NEW

• Released: May 2026
• Opus 4: New flagship — tops SWE-bench & GPQA, best-in-class coding
• Sonnet 4: Near Opus 3.5 performance at Sonnet pricing
• Both feature extended thinking & tool use as first-class features
• 200K context window, excellent vision capabilities
• Opus 4 pricing: $15.00/$75.00 | Sonnet 4: $3.00/$15.00 per 1M in/out

Google — Gemini 2.5 Pro & Flash FLASH NEW

• Pro: Tops LMArena leaderboard, #1 reasoning model, 1M context (2M coming)
• Flash: Released April 2026 — cheap reasoning model, 1M context
• Best multimodal model — native video/audio/image understanding
• Built-in thinking mode, code execution, Google Search grounding
• Pro pricing: $1.25/$10.00 | Flash: $0.15/$0.60 per 1M in/out
• Free tier available through Google AI Studio

Meta — Llama 4 (Scout, Maverick) NEW

• Scout: 109B MoE, 10M token context — longest of any open model
• Maverick: 400B MoE, competitive with GPT-4o
• Open weights, self-hostable. Vision still maturing
• Behemoth (288B active) still training
• Hosted pricing: ~$0.20-0.80/1M tokens
• ⚠️ Benchmark controversy: Meta submitted dev-only variant to LMArena

Alibaba — Qwen 3 NEW

• Released: April 29, 2026 — full family 0.6B to 235B (MoE, 22B active)
• Open weights (Apache 2.0), competitive with Claude Sonnet 4
• Hybrid thinking mode (toggle fast vs reasoning)
• Qwen3-VL (multimodal) expected soon
• Hosted pricing: ~$0.50-1.50/1M tokens for 235B

DeepSeek

• R1 (Jan 2026): Best value reasoning model, open-weight (MIT)
• V3-0324 (March update): Improved coding, competitive with GPT-4o
• Pricing: R1 $0.55/$2.19 | V3 $0.27/$1.10 per 1M in/out
• No native vision — limitation for image tasks
• R2 rumored but not released

Mistral — Medium 3

• Released April 2026, competitive with GPT-4o/Sonnet 3.5
• Pricing: $0.40/$2.00 per 1M in/out
• Unique: available for on-premises deployment (proprietary but self-hostable)
• Small 3.1 (open-weight) adds vision capabilities

Meshy v4 NEW

• Major release (April 2026): improved quad mesh topology, PBR texture generation
• API v4 with Python/Node SDKs, batch processing for production-scale assets
• Blender addon updated for v4 API
• Pricing: Free tier + Pro $20/mo + API at $0.05/model

Tripo3D V2.5 UPDATED

• Multi-view generation with improved consistency
• "TripoSG" — sparse-guided generation via sketch/depth
• Production API with webhook callbacks, ComfyUI nodes
• Export: GLB, OBJ, FBX, USDZ with PBR textures
• Pricing: Free 10/mo, Pro $15/mo

Rodin Genie 2.0 UPDATED

• Full-body avatar generation from single photo
• Unreal Engine plugin with blendshape export
• New "Studio" mode for posing and expression editing

🌐 Gaussian Splatting & Neural Rendering

• RealityCapture v2025.1: Experimental Gaussian splat export — directly relevant to your workflow
• Polycam: Full splat capture pipeline (iPhone LiDAR → Unity/UE)
• 4D Gaussian Splatting: NVIDIA research — dynamic temporal splats from video
• Compression: 10-50x size reduction (Splatfacto/nerfstudio)
• UE5.4: Community Niagara-based renderer; no official Epic support yet

🖌️ AI Texturing Tools

• Substance 3D Painter: Firefly text-to-texture with PBR channel generation, seamless tiles
• Layer AI: Project-wide style consistency, batch texturing, USD/glTF export
• Polyhive: API-driven AI texturing for pipeline integration
• Meshy v4 Texture: Improved PBR with better seam handling

🔓 Open Source 3D AI

• TRELLIS (Microsoft Research): Structured 3D generation from images — released April 2026
• Stable Fast 3D (Stability AI): Fast local image-to-3D, ComfyUI nodes available
• ComfyUI 3D Nodes: Growing ecosystem for Tripo, Meshy, Stable Fast 3D

Houdini 20.5 NEW

• Copilot (Expanded): AI assistant for Python/VEX — now covers SOPs, VOPs, DOPs
• ML Deformer SOP (experimental): Learned character deformations
• Neural Render SOP (experimental): AI-accelerated render preview
• PDG AI Integration: TOP networks can call cloud AI APIs (Meshy, Tripo) as tasks
• Community: Gaussian splat .ply importer HDA (Orbolt)

Unreal Engine 5.4 NEW

• ML Deformer v2: Better quality, lower latency for real-time character deformation
• Neural Network Module: More inference operators for custom ML in-engine
• ML-accelerated Lumen GI sampling for better performance
• AI Virtual Production: Camera tracking refinement + LED wall calibration
• RealityScan → UE: Direct Nanite mesh export with auto LODs

Nuke 16 NEW

• NukeX Copilot: AI assistant for node graph navigation & expression writing
• ML Roto Node (Improved): Better edge refinement & temporal coherence
• AI Color Match: ML-based color matching between shots
• Smart Vector Distort: ML-driven vector generation for warping/aligning

Silhouette 2025 NEW

• AI Roto v3: Hair & transparent edges, multi-object tracking, temporal stabilization
• AI Paint: Content-aware ML paint & clone tool
• Nuke Inviso plugin: Full AI roto data exchange with Nuke

After Effects 25.2 UPDATED

• Roto Brush 3 (preview): Next-gen ML rotoscoping
• Firefly Generative Fill: Text-prompt-based fill for regions
• Content-Aware Fill improved with AI-driven fill

Wonder Studio (Adobe) UPDATED

• Improved CG character compositing with AI lighting match
• Better body tracking for complex poses
• Now exports to After Effects with proper layer structure

🎯 For Your VFX Workflow

Use Case	Best Choice
Background replacement	Seedance 2.0
Multi-shot character consistency	Runway Gen-4 or Kling Master
VFX-style transformations	Pika Effects
Complex pipeline coding	Claude Opus 4
Daily driver (cost/perf)	Claude Sonnet 4 or GPT-4.1
Vision/video analysis	Gemini 2.5 Pro
Cheapest quality	Gemini 2.5 Flash ($0.15/1M in)
3D asset generation	Meshy v4 or Tripo3D
AI roto (Nuke)	Silhouette 2025 AI Roto v3 or Nuke 16 ML Roto
Gaussian splatting	RealityCapture v2025.1 + Polycam

🔮 Coming Soon

• GPT-5 — OpenAI's next major model (no date)
• Qwen3-VL — Multimodal Qwen3 (expected mid-2026)
• DeepSeek-R2 — Next reasoning model (rumored)
• Llama 4 Behemoth — Meta's 288B teacher model (in training)
• Runway Gen-4 Turbo — Faster Gen-4 variant (expected)