Model library
Every AI video model, one honest index.
Text-to-video and image-to-video models we track and battle-test with identical prompts. Specs from official documentation, prices from real API bills — not press releases. Updated July 2026.
Deep dives
Seedance 2.0
Best reference handling in the field — 9 images + 3 videos + 3 audio per generation, native lip-sync, 15 s clips. Our default pipeline model.
Full review: pricing, API, prompting →
Veo 3.1
The cinematic polish leader. Standard tier for hero shots, Fast tier is the best price/quality deal of 2026 at ~$0.96 per 8 s 1080p clip.
Full review: pricing, API, prompting →
Full index — july 2026
Seedance 2.5
previewAnnounced June 2026: 30-second coherent clips and up to 50 multimodal references. Early rollout — we'll battle-test it as soon as API access lands.
deep dive in progress
Kling 3.0
activeCheapest usable generation and the resolution leader, with a free tier. Great for drafts and volume content; limited references push retry costs up on brand work.
deep dive in progress
Sora 2
deprecatedStill the physics and world-consistency leader, but OpenAI has set a shutdown date of September 24, 2026. Do not build new pipelines on it.
deep dive in progress
Hailuo 02
activeThe budget workhorse — we generated our hero creature clip with it. Strong motion and character work for the price; no audio, fewer format options.
deep dive in progress
Runway Gen-4
activeModel plus a full editing suite: motion brush, camera controls, inpainting. Pick it when you need creative control inside one product rather than raw API output.
deep dive in progress
Luma Dream Machine (Ray)
activeFluid, dreamy motion and good image-to-video from a single frame. A solid alternate take when Seedance or Kling interpret a brief too literally.
deep dive in progress
Pika 2.5
activeBuilt for social-first effects (Pikaffects, scene swaps) rather than realism. Fast and cheap for meme-adjacent short-form content.
deep dive in progress
Wan 2.5
activeAlibaba's video family with strong open-weights heritage. Competitive quality at aggressive pricing; strongest in the value tier alongside Hailuo.
deep dive in progress
Hunyuan Video
activeThe open-source reference point. Run it on your own GPUs for full control and zero per-clip fees — at the cost of setup work and lower ceiling quality.
deep dive in progress
LTX-2
activeSpeed-and-cost champion of the open models — near-realtime drafts. Quality trails the frontier, but for iteration loops it's unbeatable per dollar.
deep dive in progress
Veo 3 (previous gen)
legacySuperseded by Veo 3.1, which is better and cheaper across every tier. Only relevant if you have prompts tuned against it — migrate.
deep dive in progress
Every profile gets promoted to a full review once we've run the model through real battles: identical prompts, logged costs, published failures. See how we test.