Skip to content
Model infoModelsSnapshot: 15 Oct 2026

Sync Labs / Sync lip-sync

Lipsync 2 Pro

Lip-sync model with diffusion-based detail enhancement and optional frame-analysis controls.

availableClosed weightsVideo editingLip synchronization
Official documentation (opens in a new tab)Reviewed 14 Oct 2026
10 of 22 specification fields collected
Output resolution512×512 generated face region with enhanced detail preservationSee specification notes below
Video input limitCredit plans: Hobby 1 minute; Starter 2 minutes; Pro 5 minutes; Team 10 minutes. Free 20 seconds except Sync 3 15 secondsSee specification notes below
Input billingSee ratesSync Labs
Output billingSee ratesPublished units and conditions below

Specifications

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit
Not collected

No exact model-specific published contract established; missing information does not mean unsupported.

Maximum output
Not collected

No exact model-specific published contract established; missing information does not mean unsupported.

Context caching
Not collected

No exact model-specific published contract established; missing information does not mean unsupported.

Reasoning
Not collected

No exact model-specific published contract established; missing information does not mean unsupported.

Tool calling
Not collected

No exact model-specific published contract established; missing information does not mean unsupported.

Structured output
Not collected

No exact model-specific published contract established; missing information does not mean unsupported.

Streaming
Not collected

No exact model-specific published contract established; missing information does not mean unsupported.

Batch processing
SupportedSource for Batch processing (opens in a new tab)

20–500 generation requests, maximum 30-second inputs; model-specific restrictions such as React 1’s 15 seconds remain in force. Requires eligible account access.

Fine-tuning
Not collected

No exact model-specific published contract established; missing information does not mean unsupported.

Parameters
Not collected

No exact model-specific published contract established; missing information does not mean unsupported.

Architecture
Independent two-second inference chunks with diffusion-based super resolution for fine facial detailsSource for Architecture (opens in a new tab)
Knowledge cutoff
Not collected

No exact model-specific published contract established; missing information does not mean unsupported.

Languages
Not collected

No exact model-specific published contract established; missing information does not mean unsupported.

Release date
Not collected

The official SDK already names this selector before the launch/changelog article. Exact first release is not established.

Input modalities
Video, AudioSource for Input modalities (opens in a new tab)

Text-to-speech inputs are synthesized by an upstream provider; the visual model is driven by audio.

Output formats
H264 videoSource for Output formats (opens in a new tab)

Output is re-encoded in an RGB pipeline; bitrate and frame rate need not match the source.

Input video formats
MP4, MOV / QT, WebM, AVISource for Input video formats (opens in a new tab)
Audio formats
WAV, MP3, OGG, FLAC, ALAC, MP4 audioSource for Audio formats (opens in a new tab)

WMA, M4A, M3A and AAC have limited support.

Video input duration
Credit plans: Hobby 1 minute; Starter 2 minutes; Pro 5 minutes; Team 10 minutes. Free 20 seconds except Sync 3 15 secondsSource for Video input duration (opens in a new tab)

Legacy plan limits are listed separately under endpoint access. Enterprise may negotiate different limits; batch inputs have a separate 30-second cap.

Output resolutions
512×512 generated face region with enhanced detail preservationSource for Output resolutions (opens in a new tab)

Face-generation dimensions, not a 512-pixel limit on the complete output video.

Generation controls
temperature 0–1 (default 0.5), optional obstruction detection and active-speaker selectionSource for Generation controls (opens in a new tab)

Access & pricing

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

API model ID
lipsync-2-pro
Access
POST https://api.sync.so/v2/generate with x-api-key; poll GET /v2/generate/{id} or receive completion webhooks. Account-available models can be listed through authenticated GET /v2/models.
Regions
Privacy policy describes direct collection in the United States and transfers to other jurisdictions; no model-specific inference residency guarantee established.
delivery
Asynchronous generation with status polling or webhooks. Status notifications and faster processing do not establish real-time streaming inference.Source for delivery (opens in a new tab)
rendering
H264 via libx264, CRF 17 and slow preset. HDR normalizes to SDR; alpha channels are removed. Original bitrate, frame rate and color appearance may change.Source for rendering (opens in a new tab)
creditPlans
Hobby $5/month: 300 credits, 1 concurrent job, 1-minute max, 500 MiB files. Starter $12: 700, 5 jobs, 2 minutes, 2.5 GiB. Pro $29: 2,000, 10 jobs, 5 minutes, 2.5 GiB. Team $99: 8,000, 20 jobs, 10 minutes, 5 GiB. Organization enrollment determines billing system.Source for creditPlans (opens in a new tab)

Model-specific limits can be lower.

legacyPlans
Hobbyist $5/month: 1 concurrent job, 1 minute. Creator $19: 3 jobs, 5 minutes. Growth $49: 6 jobs, 10 minutes. Scale $249: 15 jobs, 30 minutes. Usage billed separately; Enterprise custom.Source for legacyPlans (opens in a new tab)
topUps
Credit top-ups: Hobby $0.013, Starter $0.012, Pro $0.011, Team $0.010 per credit. Consumption value of $0.01/credit does not mean every credit costs one cent to purchase.Source for topUps (opens in a new tab)
creditExpiry
Monthly credits roll over one extra billing period. Top-ups persist while subscribed; cancellation forfeits all remaining credits, including top-ups.Source for creditExpiry (opens in a new tab)
fileLimits
Application maximum 5 GiB/file, subject to lower plan and transport limits. Credit-plan caps vary. Use upload-first assets or direct media URLs; URL inputs still must satisfy applicable limits.Source for fileLimits (opens in a new tab)
batchContract
20–500 unique requests in JSONL or JSON rows, up to 5 MB. Target turnaround 24 hours. Normal generation concurrency is bypassed, but separate active-batch limits apply. Dedicated guide says Scale or Enterprise; confirm credit-plan entitlement. No blanket batch price discount established.Source for batchContract (opens in a new tab)
privacyControls
Enterprise ZDR is contract-backed and enabled by Sync. Saved assets, generation files and account policies require the applicable agreement; no blanket default deletion deadline asserted.Source for privacyControls (opens in a new tab)
Billing basis
Legacy usage is per output frame, with published per-second examples at 25 fps. Credit plans consume credits; subscriptions, top-up purchase rates and usage consumption are separate.Source for Billing basis (opens in a new tab)
costEstimate
POST /v2/generations/estimate with positive finite duration and fps (default 30). It does not inspect media. Use actual billable duration; estimatedCredits is null for dollar-billed organizations.Source for costEstimate (opens in a new tab)
Included free allowance
Free: 3 lip-sync generations/month up to 20 seconds; at most one Sync 3 up to 15 seconds. No ongoing free generation allowance after subscribing. React 1’s 15-second cap still applies.Source for Included free allowance (opens in a new tab)
syncModes
bounce, loop, cut_off, silence, remapSource for syncModes (opens in a new tab)

Controls source video/audio length mismatch, not five separate models. Sync mode is ignored for still-image inputs.

cutOffBilling
cut_off processes and bills the shorter video/audio duration; discarded trailing content is not charged as processed video.Source for cutOffBilling (opens in a new tab)
sourceMotion
Requires natural speaking motion in the source. Static/still segments do not provide the visual cues needed for speaker-style lip generation.Source for sourceMotion (opens in a new tab)
speakerSelection
Automatic video speaker detection or manual frame/point or bounding-box selection. Availability depends on the account plan.Source for speakerSelection (opens in a new tab)
frameAnalysis
Optional reasoning_enabled analyzes and corrects frame artifacts, occlusion and extreme poses. This is not general LLM reasoning or chain-of-thought output.Source for frameAnalysis (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 14 Oct 2026
CategoryUsageRateUnit
Approximate credit consumptionCredit-billed accounts · approximately18.3 creditssecond at 25 fps
Legacy-plan usage billingHobbyist legacy usage2$0.00333generated output frame
Legacy-plan usage billingCreator legacy usage2$0.00333generated output frame
Legacy-plan usage billingGrowth legacy usage2$0.00317generated output frame
Legacy-plan usage billingScale legacy usage2$0.00267generated output frame

Rate conditions

  1. Provider publishes approximate credit consumption at 25 fps; actual lip-sync usage counts output frames. Credit purchase prices differ by plan. Not a dollar rate, not an exact per-second minimum and not an LLM token quote.
  2. Published legacy subscription-plus-usage tariff; monthly subscription separate. Cash usage plan, not the newer credit plans. Bill actual processed output frames. Displayed per-second rates on the pricing page can differ slightly because of rounding; do not assume a universal 25-fps input or a credit discount.
Pricing source (opens in a new tab)Pricing source (opens in a new tab)

Data & enterprise terms

Privacy policy permits service improvement and AI model training/fine-tuning uses. Enterprise zero-data-retention requires a contract and provider activation; it is not a default or self-serve switch. No universal fixed deletion window is established.

Read the service policy (opens in a new tab)

Caching

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

01 / INITIAL INPUTProcess & retain

The system processes eligible input and retains reusable state.

02 / LATER REQUESTCheck available state

Reuse depends on matching input, session state and the cache policy.

03 / CACHE HITReuse & generate

Reusing state can reduce repeated work. Any charges follow the billing rules below.

Sync Labs

lipsync-2-pro
Not verified

No provider prompt-cache contract established. Reusing uploaded assets, idempotent submissions or downloading an existing result is separate.

How it works
Not collected

No provider prompt-cache contract established. Reusing uploaded assets, idempotent submissions or downloading an existing result is separate.

Minimum reusable input
Not collected

No provider prompt-cache contract established. Reusing uploaded assets, idempotent submissions or downloading an existing result is separate.

Lifetime & refresh
Not collected

No provider prompt-cache contract established. Reusing uploaded assets, idempotent submissions or downloading an existing result is separate.

How to enable it
Not collected

No provider prompt-cache contract established. Reusing uploaded assets, idempotent submissions or downloading an existing result is separate.

Usage fields
Not collected

No provider prompt-cache contract established. Reusing uploaded assets, idempotent submissions or downloading an existing result is separate.

Billing rules
Not collected

No provider prompt-cache contract established. Reusing uploaded assets, idempotent submissions or downloading an existing result is separate.

Limits & data boundaries
Not collected

No provider prompt-cache contract established. Reusing uploaded assets, idempotent submissions or downloading an existing result is separate.

No verified cache rate for this endpoint. An empty rate is not a zero charge.

Caching reviewed 14 Oct 2026

What the measurements say

No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

The record behind the model

Why it is included

Official model documentation or release history identifies a Sync-developed generative lip-sync model. The public service exposes hosted inference, while the service terms reserve proprietary technology. The founders’ separate Wav2Lip open-source work does not establish that these model weights are public.

Classification evidence (opens in a new tab)

Lifecycle dates

Public launch/changelog announcement1 Sept 2025Source (opens in a new tab)
Selector present in published SDK21 Aug 2025Source (opens in a new tab)

Sources reviewed