Actual clips rendered through the pipeline - dogfooded on Trent's own voice, photo and footage.
Full cold-openFantasyTalking
249-frame single-shot, your VO, seamless · Wan2.1-14B / RunPod A100
Take 1FantasyTalking
3.5s test · natural · seed 1111
Take 2FantasyTalking
3.5s test · tighter lip-sync · seed 42
Live from the flop catalog. Each card: cost in FLOP, commercial vs research-only license, live vs preview, and what it can do.
FantasyTalkingAlibaba AMAP
portrait + audio (+ optional motion prompt); Wan2.1-14B; PROVEN on RunPod
image-to-videotalking-head-animationaudio-drivenlip-sync
Apache-2.0 (adapter + Wan2.1 base)
Hallo3Fudan
portrait + audio + scene prompt; long-clip, full-scene motion; H100-class
image-to-videotalking-head-animationaudio-drivenlip-sync
MIT (code); CogVideoX-5B weights, verify commercial
SadTalkerTencent / XJTU
portrait + audio; fast warp/3DMM, runs on modest GPU or CPU
image-to-videotalking-head-animationaudio-drivenlip-sync
Apache-2.0 (code); bundles Wav2Lip NON-COMMERCIAL
EchoMimicAnt Group
portrait + audio (+ optional landmarks/pose); SD1.5 / AnimateDiff
image-to-videotalking-head-animationaudio-drivenlip-sync
Apache-2.0 (code); weights research disclaimer, verify
SonicTencent
portrait + audio; high quality, Stable-Video-Diffusion based
image-to-videotalking-head-animationaudio-drivenlip-sync
CC BY-NC-SA 4.0 (NON-COMMERCIAL); SVD base
AniPortraitTencent Games
portrait + audio; SD1.5 / AnimateDiff, 512x512, strong identity
image-to-videotalking-head-animationaudio-drivenlip-sync
Apache-2.0 (code); SD1.5 OpenRAIL-M base, verify
LivePortraitKuaishou
portrait + driving video; near real-time, warp-based, high quality
image-to-videoface-reenactmentvideo-drivenreal-time
MIT (code); swap out InsightFace detector to stay clean
FlashPortraitFudan + MSRA
portrait + driving video; Wan2.1-14B, infinite-length ID-preserving
image-to-videoface-reenactmentvideo-driven
Apache-2.0 (code + Wan2.1 base)
FaceFusionHenry Ruhs
source face + target video; headless onnxruntime; pin hyperswap (not inswapper)
image-to-videoface-swap
OpenRAIL-AS (code); default hyperswap weights commercial-ok
GHOST (sber-swap)Sber AI
source face + target video; one-shot, no per-face training
image-to-videoface-swap
Apache-2.0 (code + Sber weights)
SimSwapSJTU
source face + target image/video; one-shot 224/512
image-to-videoface-swap
CC BY-NC 4.0 (NON-COMMERCIAL)
Deep-Live-Camhacksider
source face + target video, or real-time webcam swap
image-to-videoface-swapreal-time
AGPL-3.0 (code); inswapper_128 weights NON-COMMERCIAL
DeepFaceLabiperov
cinematic quality; TRAINS per face-pair (hours-days), not on-demand
image-to-videoface-swap
GPL-3.0 (code); you train your own weights
Character swap / body
The WHOLE person - face, hair, body, clothes. Replace the person in a video, or animate a reference person from a photo + motion.
Wan-AnimateAlibaba Wan
replace the WHOLE person in a video with your character (Mix mode), keeps motion + expression; Wan2.x-based
image-to-videocharacter-swapvideo-driven
Apache-2.0 (Wan)
Wan VACEAlibaba Wan
unified video edit: actor replacement, masked-region repaint, background swap; Wan2.1-based
image-to-videocharacter-swapvideo-driven
Apache-2.0 (Wan)
MimicMotionTencent
reference person photo + pose video -> full-body motion clip; SVD-based, ~16GB VRAM
image-to-videocharacter-swappose-driven
Apache-2.0 (code); Stability SVD base = NON-COMMERCIAL
MusePose (Animate Anyone)Tencent Music / Lyra Lab
reference person + driving pose -> animated; MusePose = the open Animate-Anyone reimpl
image-to-videocharacter-swappose-driven
MIT (code); SD1.5 + AnimateDiff OpenRAIL-M base
UniAnimateAlibaba + HUST
reference person + pose; efficient long-video (1 min) human animation
image-to-videocharacter-swappose-driven
NON-COMMERCIAL (research only); SD2.1 base
Animate-XAnt Group
universal character animation (robust to stylized/anthropomorphic refs), pose-driven
image-to-videocharacter-swappose-driven
Apache-2.0 (code); SD2.1 OpenRAIL-M base
EchoMimic V2Ant Group
semi-body avatar, audio + pose driven (talks and gestures); ~16GB VRAM
image-to-videocharacter-swappose-drivenaudio-driven
Apache-2.0 (code + weights); SD1.5 OpenRAIL-M base
Wan 2.2 i2v A14BAlibaba Wan
image-to-video, ~5s, needs init_image
image-to-video
Apache-2.0
Wan-Alpha (transparent)Wan + community DoRA
RGBA transparent video, ~5s
image-to-videotransparent-output
Apache-2.0
Wan 2.2 FLF (first-last frame)Alibaba Wan
first + last frame conditioning, ~5s; needs init_image + last_frame
image-to-video
Apache-2.0
Veo 3.1 (image-to-video)Google
8s 1280x720 with audio; needs init_image
image-to-video
Proprietary (Vertex API)
Veo 3.1 FLF (first-last frame)Google
first + last frame interpolation, 8s; needs init_image + last_frame. Frame-lock verified 99.6%/99.4%
image-to-video
Proprietary (Vertex API)
Veo 3.1 FastGoogle
faster/cheaper Veo 3.1; also accepts last_frame
image-to-video
Proprietary (Vertex API)
Veo 3.1 (text-to-video)Google
prompt-only video, 8s 1280x720 with audio; no init_image needed
image-to-video
Proprietary (Vertex API)
Stable Diffusion XLStability AI
1024x1024 (portrait/landscape ok)
text-to-imagepainterlyimage-to-image
CreativeML OpenRAIL++
Juggernaut XLRunDiffusion
1024x1536
text-to-imagephotorealisticpainterlyimage-to-image
CreativeML OpenRAIL++
DreamShaper XLLykon
1024x1536
text-to-imagepainterlyimage-to-image
CreativeML OpenRAIL++
RealVisXL V5SG161222
1024x1536
text-to-imagephotorealisticimage-to-image
CreativeML OpenRAIL++
ZavyChroma XLZavy
1024x1536
text-to-imagepainterlyimage-to-image
CreativeML OpenRAIL++
EpicRealism XLepinikion
1024x1536
text-to-imagephotorealisticimage-to-image
CreativeML OpenRAIL++
Playground v2.5Playground AI
1024x1024
text-to-imagepainterlyimage-to-image
Playground v2.5 Community
KolorsKuaishou
1024x1024
text-to-imagepainterly
Apache-2.0
FLUX.1 devBlack Forest Labs
1024x1536
text-to-imagepainterly
FLUX.1-dev Non-Commercial
Qwen-ImageAlibaba Qwen
1024x1536
text-to-imagepainterly
Apache-2.0
SANA 1.6BNVIDIA
1024x1024 (fast)
text-to-image
NSCLv2
SANA Sprint 1.6BNVIDIA
1024x1024 (2-step distilled; ~5x faster; img2img)
text-to-imageimage-to-image
NSCLv2
PixArt-SigmaPixArt-alpha
1024x1024
text-to-imagepainterly
OpenRAIL++
AbsoluteReality (SD1.5)Lykon
512x768
text-to-imagephotorealisticimage-to-image
CreativeML OpenRAIL-M
Realistic Vision V6 (SD1.5)SG161222
512x768
text-to-imagephotorealisticimage-to-image
CreativeML OpenRAIL-M
DreamShaper 8 (SD1.5)Lykon
512x768
text-to-imagepainterlyimage-to-image
CreativeML OpenRAIL-M
Stable Diffusion 3.5 LargeStability AI
1024x1024, gated weights
text-to-imagepainterly
Stability AI Community
Stable Image UltraStability AI
hosted API, top-tier 1MP+
text-to-imagepainterly
Proprietary (Stability API)
HiDream-I1 FullHiDream AI
1024x1024, needs a Llama text encoder
text-to-imagepainterly
MIT
HunyuanImage 3.0Tencent
native multimodal, 1024x1024+
text-to-imagepainterly
Tencent Hunyuan Community
Z-ImageTongyi Lab (Alibaba)
6B efficient, 1024x1024
text-to-imagepainterly
Apache-2.0
Z-Image TurboTongyi Lab (Alibaba)
few-step distilled, fast 1024x1024
text-to-image
Apache-2.0
GPT Image 1.5 (transparent)OpenAI
text-to-image with transparent-background RGBA PNG output
text-to-imagetransparent-outputpainterly
Proprietary (OpenAI API)
Imagen 4Google
high-fidelity text-to-image
text-to-image
Proprietary (Vertex API)
Retro Diffusion (RD Pro)Retro Diffusion
true pixel art, up to 256px
text-to-imagepixel-art
Commercial (RD API)
Pixel Art XL (SDXL LoRA)nerijs
1024x1024 + kCentroid downscale
text-to-imagepixel-art
CreativeML OpenRAIL++
Pixel Art Redmond (SDXL)ArtificialGuyBr
1024x1024
text-to-imagepixel-art
CreativeML OpenRAIL++
Nano Banana (Gemini 2.5 Flash Image)Google
instruction image edit + generation; needs init_image for edits
instruction-editimage-to-image
Proprietary (Vertex / Gemini API)
GPT Image 2OpenAI
top-tier image gen + instruction edit, up to 2K
instruction-editimage-to-image
Proprietary (OpenAI API)
Qwen-Image-EditAlibaba Qwen
instruction edit, needs init_image
instruction-editimage-to-image
Apache-2.0
FLUX.1 Kontext devBlack Forest Labs
instruction edit, needs init_image
instruction-editimage-to-image
FLUX.1 Non-Commercial
Step1X-EditStepFun
instruction edit, needs init_image
instruction-editimage-to-image
Apache-2.0
InstructPix2PixTim Brooks et al.
instruction edit, needs init_image
instruction-editimage-to-image
MIT
ACE-Step v1 3.5BACE Studio / StepFun
music, up to ~4 min, 48kHz stereo
audio-generationmusic
Apache-2.0
Stable Audio OpenStability AI
sound effects, up to ~47s
audio-generationsound-effects
Stability Community
AudioLDM (m-full)Haohe Liu / CVSSP
text-to-audio: sfx, ambience, short music; ~10s
audio-generationsound-effects
CC BY-NC-SA 4.0 (non-commercial)
TangoFluxDeclare Lab
sound effects, ~10s
audio-generationsound-effects
Non-Commercial (research)
Llama 3.2 3BMeta
128k context, fast
text-generationchat
Llama 3.2 Community
Llama 3.3 70BMeta
128k context, flagship open
text-generationchat
Llama 3.3 Community
Llama 3.1 8BMeta
128k context, balanced
text-generationchat
Llama 3.1 Community
Qwen2.5 72BAlibaba Qwen
128k context, top open general
text-generationchat
Qwen License
Qwen2.5 Coder 32BAlibaba Qwen
code-specialized
text-generationcode-generation
Apache-2.0
DeepSeek-R1 70BDeepSeek
reasoning / chain-of-thought
text-generationreasoning
MIT
Mistral Small 3 24BMistral AI
32k context
text-generationchat
Apache-2.0
Gemma 3 4B (vision)Google
multimodal text + vision (i2text); self-hosted vLLM on RunPod
text-generationchatimage-understanding
Gemma
Phi-4 14BMicrosoft
small but strong
text-generationchat
MIT
Llama 3.2 Vision 11BMeta
image understanding (i2text); self-hosted vLLM on RunPod
text-generationimage-understanding
Llama 3.2 Community
Qwen2.5-VL 7BAlibaba Qwen
image + document understanding (i2text); self-hosted vLLM on RunPod
text-generationimage-understanding
Apache-2.0
LLaVA 1.6 (Mistral 7B)LLaVA
lightweight vision-language (i2text); self-hosted vLLM on RunPod
text-generationimage-understanding
Apache-2.0
Claude Opus 4.8Anthropic
frontier reasoning + vision, 200k+ context
text-generationchatreasoningimage-understanding
Proprietary (Anthropic API)
Claude Fable 5Anthropic
frontier creative writing
text-generationchatcreative-writing
Proprietary (Anthropic API)
Claude Sonnet 5Anthropic
balanced frontier, fast
text-generationchatreasoningimage-understanding
Proprietary (Anthropic API)
Claude Haiku 4.5Anthropic
fast + cheap frontier
text-generationchat
Proprietary (Anthropic API)
GPT-4oOpenAI
multimodal frontier + vision (text + image-to-text)
text-generationchatimage-understanding
Proprietary (OpenAI API)
Gemini 2.5 ProGoogle
long-context multimodal, 1M tokens
text-generationchatimage-understanding
Proprietary (Google API)