Skip to content
Model infoModelsSnapshot: 15 Oct 2026

Decart / Lucy

Lucy 2.5

Edits video from text, reference images or both, through realtime and queued APIs.

Video generation
availableClosed weightsVideo editing
Official documentation (opens in a new tab)Reviewed 15 Oct 2026
6 of 20 specification fields collected
Output resolution720pSee specification notes below
Clip durationNot collectedAwaiting source review
Input billingSee ratesDecart realtime API
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

Not established for this exact model and API. Media generation does not imply language-model token limits or capabilities.

Maximum input tokens
Not collected

Not established for this exact model and API. Media generation does not imply language-model token limits or capabilities.

Maximum output
Not collected

Not established for this exact model and API. Media generation does not imply language-model token limits or capabilities.

Reasoning
Not collected

Not established for this exact model and API. Media generation does not imply language-model token limits or capabilities.

Tool calling
Not collected

Not established for this exact model and API. Media generation does not imply language-model token limits or capabilities.

Structured output
Not collected

Not established for this exact model and API. Media generation does not imply language-model token limits or capabilities.

Streaming
SupportedSource for Streaming (opens in a new tab)

Realtime video over WebRTC; not streamed language tokens.

Batch processing
Not collected

Not established for this exact model and API. Media generation does not imply language-model token limits or capabilities.

Fine-tuning
Not collected

Not established for this exact model and API. Media generation does not imply language-model token limits or capabilities.

Knowledge cutoff
Not collected

Not established for this exact model and API. Media generation does not imply language-model token limits or capabilities.

Release date
Not collected

Not established for this exact model and API. Media generation does not imply language-model token limits or capabilities.

Parameters
Not collected

Not established for this exact model and API. Media generation does not imply language-model token limits or capabilities.

Architecture
Not collected

Not established for this exact model and API. Media generation does not imply language-model token limits or capabilities.

Languages
Not collected

Not established for this exact model and API. Media generation does not imply language-model token limits or capabilities.

Context caching
Not collected

Not established for this exact model and API. Media generation does not imply language-model token limits or capabilities.

Input modalities
Text, Image, VideoSource for Input modalities (opens in a new tab)
Output modalities
VideoSource for Output modalities (opens in a new tab)
Output resolutions
720pSource for Output resolutions (opens in a new tab)

Exact shapes and transport-specific constraints are recorded under each endpoint.

Release window
2026-07Source for Release window (opens in a new tab)

Month-level provider announcement; no exact release day inferred.

Model aliases
lucy-latestSource for Model aliases (opens in a new tab)

Current server-side alias mapping as reviewed; aliases are not additional models and may move.

Access & pricing

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

Decart realtime API

Provider source (opens in a new tab)
API model ID
lucy-2.5
Access
Native realtime session via SDK/WebRTC. Account and API key required; no session started.
Regions
Not established
API endpoint URL
wss://api.decart.ai/v1/stream?model=lucy-2.5Source for API endpoint URL (opens in a new tab)

WebSocket signaling with WebRTC media. api3.decart.ai remains a compatibility hostname, not another model.

Authentication
API key or short-lived client token; browser/mobile tokens should be minted by the application backend.Source for Authentication (opens in a new tab)
Live session controls
Prompt/image updates replace the whole session state; omitted values are cleared. Token expiry prevents new connections but does not terminate an active session.Source for Live session controls (opens in a new tab)
Session duration
No fixed global duration collected. Client-token expiresIn is 1–3600 seconds, default 60; maxSessionDuration is a separate optional session cap.Source for Session duration (opens in a new tab)
Required source video format
SDK capture dimensions 1280×720; portrait orientation supported by the model guide.Source for Required source video format (opens in a new tab)

Input capture configuration, not a guaranteed delivered output rate or latency.

Frame rate
SDK capture requests ideal/max 30 FPS; not an output FPS or latency guarantee.Source for Frame rate (opens in a new tab)
Generation mode
Standard, Fast (speed=fast at connect time)Source for Generation mode (opens in a new tab)

Fast changes throughput and price; changing it requires reconnecting. It is not another model version.

Billing basis
Active generation time, measured in seconds. Realtime output is not billed as language tokens.Source for Billing basis (opens in a new tab)
Original generation context
Self-anchoring enabled by default; self_anchor=false disables it at connection time. A scene cut may leave an unsuitable prior-frame anchor.Source for Original generation context (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 15 Oct 2026
CategoryUsageRateUnit
Media generationRealtime standard1$0.02active generation second
Media generationRealtime fast2$0.04active generation second

Rate conditions

  1. 720p realtime; active generation time.
  2. speed=fast; twice the standard rate. Current docs and SDK support this exact version.
Pricing source (opens in a new tab)

Data & enterprise terms

Reviewed API terms permit input/output use for service improvement and training, and do not treat them as confidential. Enterprise-specific overrides were not reviewed.

Read the service policy (opens in a new tab)

Decart video jobs API

Provider source (opens in a new tab)
API model ID
lucy-2.5
Access
Native asynchronous recorded-video job API. Account and API key required; no inference job submitted.
Regions
Not established
API endpoint URL
POST https://api.decart.ai/v1/jobs/lucy-2.5Source for API endpoint URL (opens in a new tab)

Multipart request; model-specific path confirmed in OpenAPI.

Authentication
X-API-KEY request header.Source for Authentication (opens in a new tab)
Processing workflow
Submit, then GET /v1/jobs/{job_id}; download GET /v1/jobs/{job_id}/content after completion. Status: pending, processing, completed or failed.Source for Processing workflow (opens in a new tab)

Asynchronous video processing, not a discounted language-model Batch API.

Input file formats
MP4 (H.264 or VP8)Source for Input file formats (opens in a new tab)
Input video limits
Maximum 200 MB per input video.Source for Input video limits (opens in a new tab)
Video input duration
No duration cap documented, subject to the 200 MB file limit.Source for Video input duration (opens in a new tab)
Output pixel presets
1280×720, 720×1280Source for Output pixel presets (opens in a new tab)
Reference guidance
Optional reference image and optional text prompt; missing text defaults to empty string in REST.Source for Reference guidance (opens in a new tab)
Random seed
0–4294967295; optional.Source for Random seed (opens in a new tab)
Prompt expansion
enhance_prompt defaults to true in REST. Restyle SDK applies additional reference-image restrictions.Source for Prompt expansion (opens in a new tab)
Billing basis
Per generated video second, not live session duration.Source for Billing basis (opens in a new tab)
Original generation context
self_anchor defaults to true; false disables latent self-anchoring for this job.Source for Original generation context (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 15 Oct 2026
CategoryUsageRateUnit
Media generationRecorded video 720p1$0.04generated video second

Rate conditions

  1. Rate applies to generated output duration. Queued video jobs do not establish prompt-cache support or a separate discounted Batch tariff.
Pricing source (opens in a new tab)

Data & enterprise terms

Reviewed API terms permit input/output use for service improvement and training, and do not treat them as confidential. Enterprise-specific overrides were not reviewed.

Read the service policy (opens in a new tab)

Fal realtime

Provider source (opens in a new tab)
API model ID
decart/lucy-2-5/realtime
Access
Listed in the current Fal catalog. Use a Fal API key on the backend to mint short-lived realtime tokens for the client. No authenticated inference was performed.
Regions
Not established
API endpoint URL
fal.realtime.connect("decart/lucy-2-5/realtime")Source for API endpoint URL (opens in a new tab)
Streaming
SupportedSource for Streaming (opens in a new tab)

Realtime video transformation, not language-token streaming.

Authentication
Keep FAL_KEY on the backend; the browser receives a short-lived JWT through tokenProvider. The example token lifetime is not a model session-duration limit.Source for Authentication (opens in a new tab)
Model controls
enable_prompt_expansion, image_url, prompt, reference_image_urlSource for Model controls (opens in a new tab)
Required request fields
No form fields are marked required by the JSON schema. A live media/signaling session is still needed; an empty object alone is not an inference workload.Source for Required request fields (opens in a new tab)
Reference image syntax
reference_image_url accepts a data URI, JPEG/PNG/WebP and minimum 512×512 images. The image_url playground field is unused in WebRTC mode.Source for Reference image syntax (opens in a new tab)
Reference guidance
A reference image is used for character swaps or virtual try-on; other editing operations may use a prompt alone.Source for Reference guidance (opens in a new tab)
Prompt length limit
Not collected

No prompt maxLength, reference byte ceiling or session duration limit is provided in this host schema.

Published request quota
Not collected

No numeric request/concurrency quota collected for this Fal route.

Frame rate
Not collected

Marketing describes realtime performance; no numerical output FPS or latency service guarantee is established by the schema.

Endpoint availability
Current Fal catalog advertises this route.Source for Endpoint availability (opens in a new tab)
Prompt expansion
SupportedSource for Prompt expansion (opens in a new tab)

enable_prompt_expansion defaults true.

Input media transport
Fal relays signaling; video media travels directly between the browser and Decart over WebRTC.Source for Input media transport (opens in a new tab)
Result metadata
iceServers, error, candidate, sdp, typeSource for Result metadata (opens in a new tab)

Signaling output fields, not a downloadable generated-video JSON schema.

Published price
Conflicting published prices: endpoint reference lists USD 0.02 per second; the Lucy 2.5 landing page lists USD 0.04 per video second. Confirm the account quote before estimating this route.Source for Published price (opens in a new tab)

See the separately linked landing-page price below.

Pricing discrepancy
Fal Lucy 2.5 landing page: USD 0.04 per second. This conflicts with the USD 0.02 endpoint-reference price; no documented tier or fast-mode distinction reconciles them.Source for Pricing discrepancy (opens in a new tab)
Billing basis
Not collected

Conflicting official prices prevent a single verified tariff. Do not sum the two numbers, treat them as a range, or import native fast-mode pricing.

Endpoint pricing has not been collected. Check the provider for applicable rates.

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

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.

Decart realtime API

lucy-2.5
Not verified

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

How it works
Not collected

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

Minimum reusable input
Not collected

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

Lifetime & refresh
Not collected

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

How to enable it
Not collected

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

Usage fields
Not collected

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

Billing rules
Not collected

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

Limits & data boundaries
Not collected

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

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

Caching reviewed 15 Oct 2026

Decart video jobs API

lucy-2.5
Not verified

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

How it works
Not collected

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

Minimum reusable input
Not collected

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

Lifetime & refresh
Not collected

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

How to enable it
Not collected

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

Usage fields
Not collected

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

Billing rules
Not collected

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

Limits & data boundaries
Not collected

No model prompt-cache contract or cache discount established. Upload reuse, client-token expiry and stateful video context are separate mechanisms.

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

Caching reviewed 15 Oct 2026

Fal realtime

decart/lucy-2-5/realtime
Not verified

No inference cache mechanism, retention window or cache discount is published for this route. Stateful streaming and client authentication tokens are not prompt caching.

How it works
Not collected

No inference cache mechanism, retention window or cache discount is published for this route. Stateful streaming and client authentication tokens are not prompt caching.

Minimum reusable input
Not collected

No inference cache mechanism, retention window or cache discount is published for this route. Stateful streaming and client authentication tokens are not prompt caching.

Lifetime & refresh
Not collected

No inference cache mechanism, retention window or cache discount is published for this route. Stateful streaming and client authentication tokens are not prompt caching.

How to enable it
Not collected

No inference cache mechanism, retention window or cache discount is published for this route. Stateful streaming and client authentication tokens are not prompt caching.

Usage fields
Not collected

No inference cache mechanism, retention window or cache discount is published for this route. Stateful streaming and client authentication tokens are not prompt caching.

Billing rules
Not collected

No inference cache mechanism, retention window or cache discount is published for this route. Stateful streaming and client authentication tokens are not prompt caching.

Limits & data boundaries
Not collected

No inference cache mechanism, retention window or cache discount is published for this route. Stateful streaming and client authentication tokens are not prompt caching.

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

Caching reviewed 15 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

The provider names this hosted model, retains rights in its models and limits access through its API. No matching checkpoint appears in the reviewed official public-weight inventory. Closed-weight classification is an inference from these sources; it is not applied to Lucy Edit Dev releases or public SDK code.

Classification evidence (opens in a new tab)

Sources reviewed