Four APIs. One integration.
The main agentic MCP server. Pass a video and a creative intent — Catalyst understands the scene, builds the plan, and orchestrates the full audio session end to end.
Four ways in.
Use what you need.
Give it intent. It plans, you approve, it builds.
The full agentic pipeline. Upload a video, describe your creative direction, and Catalyst's Director Agent analyzes the scene, designs the complete audio layer, and returns an implementation plan before generating a single file. You review. You approve. Then the audio is built — stems, master, and session manifest delivered together.
Inside the Agentic Pipeline.
Every agentic session runs the same four phases. Nothing skips the queue.
Empowering sound design teams of all sizes to ship at scale
Production-grade multimodal AI pipeline benchmarked for high-throughput video scoring.
Real-time, multi-node inference for cinematic Agentic Soundtracks
Autonomous scene reasoning & approval-gated blueprint generation.
Lossless 24-bit/48kHz stem synthesis running on Catalyst with 10–15 ms latency.
Multi-track neural synthesis with sub-frame musical alignment.
ML-driven Indian classical & folk composition
Powering visual sound design generation at scale
“We’re actively saving 2 sound designers’ worth of manual timeline editing”
3x latency decrease for automated DAW timeline scoring
“Catalyst makes it easy to orchestrate 100s of audio layers in parallel, delivering lossless stems, masters, and frame-accurate JSON manifests in seconds.”
Autonomous taste.
Built in cultural alignment.
Generating audio is solved. Generating audio that is right for this moment, in this culture, is not.
Every piece of audio Catalyst generates passes through the Indian Cultural Verifier before it is returned. The Verifier is not a filter — it is a judge. It evaluates whether the audio is emotionally correct for the scene, culturally appropriate for the context, and coherent across the full duration of the session.
The Verifier is fine-tuned on preference data from Indian film composers, sound designers, and regional music practitioners. It improves continuously as human editors interact with Catalyst's output — every override is a signal. Every correction makes the Verifier smarter.
Our audio is not proprietary. Our taste is. And our taste gets better for free every time a user disagrees with us.
Built for Production.
Reliability and speed are not features. They are the baseline.
Uptime target across all production endpoints.
Visual intelligence analysis for videos up to 10 minutes.
Standalone audio generation for outputs up to 60 seconds.
Full implementation plan returned and ready for approval.
Integrate in minutes via MCP or REST.
One integration. Any MCP client or plain REST. Your LLM, your stack, your infrastructure.
Connect via the official MCP ClientSession to orchestrate perception, planning, and audio engineering.
# Connect via MCP Protocol
from mcp import ClientSession
async with ClientSession() as session:
await session.initialize()
# Analyze video
result = await session.call_tool(
"sonic_analyze",
{"video_url": "https://...", "depth": "full"}
)
# Generate audio session
plan = await session.call_tool(
"catalyst_plan",
{"video_url": "https://...", "intent": "cinematic score"}
)Plain HTTP endpoints for asynchronous batch video processing, webhook events, and serverless pipelines.
# Plain REST — no MCP client needed
curl -X POST https://api.falcorelabs.com/v1/sonic/analyze \
-H "X-API-Key: YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"video_url": "https://...", "depth": "full"}'| Tool | Description |
|---|---|
| sonic_analyze | Submit video for visual intelligence analysis |
| catalyst_plan | Trigger Director Agent and return implementation plan |
| catalyst_approve | Approve plan and start audio engineering |
| catalyst_session | Retrieve stems, master, and session manifest |
| composer_generate | Direct audio generation from prompt and parameters |
| wave_generate | Music generation with tier and cultural context selection |
| catalyst_context | Retrieve full session state and history |
| sonic_search | Search asset library by text query and filters |
Founding
access is
open.
We are opening a limited number of founding integrations for creative platforms who want to embed audio and visual intelligence before the public launch. Founding partners get priority API access, dedicated onboarding, and direct input on the features we build next. If you are building a video editing tool, a content platform, or a creative workflow product — this is your invitation.
