VidScore
AI Video ModelsAI Video LeaderboardCompareBest ForCalculatorBlog
View Leaderboard
VidScore

The source of truth for AI video. Objective benchmarks, transparent data.

Platform
  • Leaderboard
  • Models
  • Compare
  • Tools
  • Cost Calculator
Resources
  • Best For Guides
  • Blog
  • Methodology
  • Editorial Policy
About

Objective benchmarks and transparent data for AI video generation. Rankings refreshed weekly.

Read our methodologyRead our editorial policy

© 2026 VidScore. Data updated April 2026.

  1. Home
  2. Models
  3. Wan 2.7

Wan 2.7

open-sourceyoutubecommercial

Alibaba · Mixture-of-Experts Diffusion Transformer · v2.7pendingStill verifying this price

Built for open source and youtube. Reasonable for serious production work.

$0.10/sec

published price via FAL.ai at 720p–1080p, not yet in the verified comparison set

Ceiling

1080p

Duration

2–15s

Price source

FAL.ai

Text-to-VideoImage-to-VideoAudioCameraLipsyncV2V

Best for

Open source and youtube

Native audio, 15s clips, and camera control support short narrative work.

Watch out for

Not verified as open weight or self-deployable despite third-party speculation about a future release

Not verified as open weight or self-deployable despite third-party speculation about a future release

Starting cost

$0.100/sec

About $0.50 for a 5-second clip at the clearest published starting tier.

Quick facts for AI assistants

Citation-ready model facts

These facts are generated from the current model record, provider verification metadata, and the page update timestamp.

Canonical URL
https://vidscore.dev/models/wan-27
Current verdict
Built for open source and youtube. Reasonable for serious production work.
Verified pricing
Published $0.10/sec via FAL.ai, not yet admitted to verified comparisons
Provider evidence
FAL.ai (Reported); Segmind (Reported)
Best fit
open-source, youtube, commercial
Primary limitation
Not verified as open weight or self-deployable despite third-party speculation about a future release
Last updated
2026-04-18

Pricing

The default rate first, with provider details one click away.

2 providers, 12 tiers

$0.100/sec

Published rate on FAL.ai · 720p–1080p

Not in the verified comparison set yet.

Audio: $0.100/sec

All provider rates

Expand for every tier, audio surcharge, and verification note.

2 providersexpand_more
FAL.aiCheapest floor

Wan 2.7 · Reported by provider · HIGH trust

Open provideropen_in_new
Text-to-Video720p–1080pAudio
$0.100/s
Image-to-Video720p–1080pAudio
$0.100/s
Reference-to-Video720p–1080pAudio
$0.100/s
Edit-Video720p–1080pAudio
$0.100/s
Checked 2026-04-18
Segmind

Wan 2.7 · Reported by provider · LOW-MEDIUM trust

Open provideropen_in_new
Text-to-Video720p
$0.63
Text-to-Video1080p
$0.94
Image-to-Video720p
$0.63
Image-to-Video1080p
$0.94
Reference-to-Video720p
$0.63
Reference-to-Video1080p
$0.94
Edit-Video720p
$0.63
Edit-Video1080p
$0.94
Checked 2026-04-14

Should you choose it?

Why pick Wan 2.7, and why pass?

Use this as a sanity check before you compare it against the closest alternatives.

thumb_upWhy choose it

  • Four modes in one architecture — t2v, i2v with frame anchoring, reference-to-video, and instruction editing
  • 27B MoE parameters (14B active) delivers premium quality at competitive $0.10/sec pricing
  • Hosted API access on FAL.ai with a flat $0.10/sec rate across text-to-video, image-to-video, reference-to-video, and edit-video
  • First/last-frame control enables precise scene transitions unavailable in most competitors
  • Native audio with voice cloning and lip-sync in reference-to-video mode
  • 1080p output at 30fps with 5 aspect ratios covers most production needs

infoWhy skip it

  • Not verified as open weight or self-deployable despite third-party speculation about a future release
  • Self-hosting requirements cannot be verified until official weights and code are released
  • AA Arena ELO of 1,188 (under Wan 2.6) may not fully reflect 2.7 improvements yet — Wan 2.7 not yet on Arena
  • No 4K output — maxes at 1080p while some competitors offer 4K
  • Newer model with less community tooling and fine-tuning ecosystem than established alternatives

Compare next

Best next comparisons

Use these when price alone is not enough. They cover the closest cost, output, and workflow tradeoffs.

All comparisons →

Wan 2.7 vs Kling v3 Pro

Est. 180/mo

Wan 2.7 for multishot scenes. Kling v3 Pro for multishot scenes.

Gap: 180 searches/mo

Search demand: 180 searches/mo

→

Quick facts

Keep the key specs and next jumps close

Resolution
720p, 1080p
Duration
2–15s
Audio
Native
Open source
No

Jump links

PricingComparePrompt guideFAQ

Quick links

open_in_newTry on FAL.aicalculateCost CalculatorleaderboardLeaderboard

Prompt playbook

auto_fix_highStart with these three moves

Scan the quick playbook first. Open the full guide only when you need deeper do/don't rules or example prompts.

  1. 1

    Structure prompts by element — describe subject, style, lighting, and composition as distinct descriptors rather than a single run-on sentence.

  2. 2

    Specify camera directions explicitly: 'camera follows,' 'smooth pan left,' 'close-up,' 'aerial descending,' 'dolly zoom' — Wan 2.7 interprets complex camera blocking.

  3. 3

    Include style keywords like 'cinematic,' 'cartoon,' 'realistic,' 'anime' to guide the model's visual treatment.

Open the full prompt guide

Extra prompt tips, best practices, common mistakes, and example prompts.

expand_more

More tips

  1. 4For first/last-frame control in image-to-video, describe the progression between the two anchor frames rather than restating what's in each image.
  2. 5For instruction-based editing, be specific: 'change the jacket from red to navy' works better than 'make it look different.'
  3. 6Leverage the 9-grid multi-image input for reference-to-video to maintain character consistency across multiple subjects.

✓ Do this

  • Follow the pattern: Who/What + Where + Movement/Activity + Camera + Style for text-to-video
  • Use endpoint anchors (first and last frame) for precise scene transitions in image-to-video mode
  • For multi-character scenes, use multi-reference grids with distinct visual identities per character
  • Keep environmental context rich: lighting conditions, time of day, weather, atmosphere
  • Use the edit-video mode for iterative refinement — cheaper than regenerating entire clips

✗ Avoid this

  • No downloadable open weights or Apache 2.0 model release verified as of April 18, 2026 — self-deployment is not available
  • No multi-shot generation in a single call — clips must be stitched manually
  • Complex multi-step sequences within a single prompt may lose coherence beyond 10 seconds
  • Text rendering in video remains unreliable
  • Very high parameter count (27B total) makes self-deployment resource-intensive when weights ship

Example prompts

Cinematic / Urban

“A bustling Tokyo street at night with neon lights reflecting on wet pavement. A young woman in a leather jacket walks toward the camera. Slow tracking shot, cinematic anamorphic lens flare, moody cyberpunk atmosphere. Rain drizzle, 1080p, 16:9.”

Product / Food

“A chef in a professional kitchen flambeing a pan of shrimp. Dramatic orange flames rise briefly. Close-up from the side, warm tungsten lighting, shallow depth of field. The chef's confident expression visible through the flames.”

Video Editing

“[Edit mode] Change the sky from overcast gray to a vivid sunset with deep orange and purple gradients. Keep all foreground elements — the lighthouse, the rocky coastline, and the waves — unchanged.”

Based on the official prompt guide →

FAQexpand_more

How much does Wan 2.7 cost?

Wan 2.7 currently has a published rate of $0.10/sec on FAL.ai (720p–1080p), but VidScore has not admitted that quote into the verified comparison set yet. A 5-second video at that tier costs about $0.50.

Where can I use Wan 2.7?

Via API on FAL.ai and Segmind.

How do I get good results with Wan 2.7?

Structure prompts by element — describe subject, style, lighting, and composition as distinct descriptors rather than a single run-on sentence. See the prompt guide below.