Reference Recreate
Use reference video to guide motion, timing, framing, cuts and performance while selectively replacing characters, products, wardrobe, backgrounds or style.
Web StudioESVYLA is a reference-driven AI creation and editing system: creators can combine reference video, characters, products, audio and optional text, control what stays locked, generate alternate takes, finish in Pro Editor and adapt masters in Campaign Director. The public API exposes the production services that are currently programmable; Studio-only capabilities are labelled clearly below.
{
"prompt": "Luxury fragrance film in rainy London",
"durationSeconds": 20,
"aspectRatio": "9:16",
"model": "smart-auto",
"idempotencyKey": "campaign-104-render-1"
}
→ ESVYLA Director plan
→ project + scenes
→ credit quote / reservation
→ asynchronous render job
→ jobId + status URL
Use reference video to guide motion, timing, framing, cuts and performance while selectively replacing characters, products, wardrobe, backgrounds or style.
Web StudioReuse Character Packs, protect identity and direct performance, expressions, wardrobe and multi-scene continuity.
Web StudioAdd multiple products, map how they are held, worn or used, and protect packaging, labels and visual identity.
Web StudioUse music or dialogue for timing, beat-aware direction, performance sync, subtitles and soundtrack-aware scene planning.
Web StudioLock what must stay consistent, change only selected elements and generate alternate takes for individual scenes.
Web StudioMove into Pro Editor for timeline, effects, color, audio and export, then adapt a master into channel-specific campaign deliverables.
Web StudioAPI scope today: the public API currently exposes Director planning, scene intelligence, model discovery, quoting, managed generation, render status, usage and pricing. Reference, Product and Character asset orchestration remains in the authenticated Web Studio until those upload and ownership contracts are exposed publicly. Director automatically routes each brief internally; those routing strategies are implementation details, not customer-facing content categories.
Turn a prompt into a directed ESVYLA project and queued render. Generation uses the account's ESVYLA Credits and returns a job ID for asynchronous tracking.
Dynamic render quotePOST /v1/videos/generateTurn a brief, script or concept into a structured treatment, timed scene plan, camera direction and continuity instructions.
£0.05 / successful callPOST /v1/director/planInspect scene complexity, reference requirements, continuity risk and quality-routing needs before generation.
£0.01 / successful callPOST /v1/intelligence/sceneEstimate required ESVYLA Credits for an existing scene plan before starting a paid render.
FreePOST /v1/videos/quoteTrack an owned render job from queue through generation and completion without holding an HTTP request open.
FreeGET /v1/videos/status?jobId=…Discover selectable and provider-gated video models programmatically. Smart Auto is recommended; ESVYLA never silently substitutes an unavailable requested model.
FreeGET /v1/modelsRead account usage, credit balance and current commercial pricing programmatically.
FreeGET /v1/usage · GET /v1/pricingLaunch catalogue: ESVYLA Smart Auto, Seedance 2.5, Google Veo 3.1 and Kling 3.0 Omni are supported generation choices. Kling becomes available only when its server-side API key and protected cost rate are configured. ESVYLA never bills a customer for an unavailable model or silently swaps it for another.
Recommended. Routes by production need, quality and protected cost.
ConnectedCinematic Seedance generation route.
ConnectedFlagship Veo 3.1 generation with native audio.
ConnectedMultimodal Kling generation with image/video references and native audio.
Server-side connectorDirector and Scene Intelligence use fixed successful-call pricing. Video generation uses ESVYLA's render quote and credit reservation system. Quote, usage and status calls are free; generation stops when the account does not have enough available credits.
curl -X POST https://eleganza-ai-music-video.web.app/v1/videos/generate \
-H "Authorization: Bearer lm_live_YOUR_KEY" \
-H "Idempotency-Key: campaign-104-render-1" \
-H "Content-Type: application/json" \
-d '{
"prompt":"Cinematic product film, neon London",
"durationSeconds":20,
"aspectRatio":"9:16",
"model":"smart-auto"
}'
# Save jobId, then poll:
GET /v1/videos/status?jobId=YOUR_JOB_ID
Keep lm_live_… keys on your server. Use a stable Idempotency-Key when retrying the same generation so network retries do not intentionally create another render. Save the returned jobId against your own user or order and poll the status endpoint.
MCP is the agent-facing surface of ESVYLA. A compatible client can inspect the platform capability map, develop a treatment, inspect scenes, discover models, estimate a render, start paid generation, check usage and follow render progress through the same ESVYLA account.
REMOTE MCP · HTTPAuthentication: Bearer lm_live_…. Generation is a paid/write action. Planning, quoting and status remain separate tools so an agent can explain what it intends to do before spending render credits.
Remote MCP URL
https://eleganza-ai-music-video.web.app/api/mcp
Tools
lumetra_capabilities
lumetra_director_plan
lumetra_scene_intelligence
lumetra_models
lumetra_video_quote
lumetra_generate_video
lumetra_render_status
lumetra_usage
Try asking
“Plan a 20-second vertical fragrance campaign.
Show me the production plan and expected render
credits, then generate it when I approve.”
Director, scene intelligence, project state, routing, billing and render workflow are exposed as ESVYLA capabilities.
The API does not imply that every underlying generation model is trained or owned by ESVYLA. Provider capabilities can evolve behind the production layer.
Developer credentials stay outside browser code. Render status is ownership-checked against the authenticated ESVYLA developer account.