create_imageLABSCO SUMMARY
A whole media pipeline behind one endpoint.
Generation, composition, editing and analysis sit on the same server, so a video can be planned, have its assets made in parallel, be assembled with transitions and ducked audio, then read back for QA without moving between services.
What it is
A hosted media production platform for agents at https://mcp.soundside.ai/mcp. Nineteen tools cover generation of images, video, audio, music, text and business artifacts, server-side video composition, editing, extraction and analysis, plus LoRA adapter training — routed across Alibaba (Wan), Grok, MiniMax, Runway, Lyria and Vertex AI.
What you get
- Text-to-image with character references for a consistent character across shots (
create_image), and text-to-video, image-to-video and video extension (create_video) - Text-to-speech, sound effects, voice cloning and voice design (
create_audio), and music from lyrics and a style prompt (create_music) - LLM completions with optional structured JSON output (
create_text), and charts, presentations, documents and diagrams through pptx, plotly, weasyprint, mermaid or gamma, with a bundle mode that produces several artifacts from one brief (create_artifact) - A server-side pipeline that enriches a plan, generates the assets in parallel and assembles them with transitions, audio ducking and overlays (
compose_video) - Editing primitives: trim, concat, crossfade, speed, loop, colour grade, burn subtitles and raw FFmpeg (
edit_video); audio mixed, replaced or padded (edit_audio); text, overlays and split screens (compose_media); Ken Burns, speed ramps, film grain and vignette (apply_effect); frames and audio tracks pulled out (extract_media) - Technical metadata, vision QA, transcription, segment detection and EDL export over a finished asset (
analyze_media) - LoRA adapters trained from your own library media, then listed, deployed, undeployed or pointed at a different checkpoint (
train_adapter,list_adapters,manage_adapter) - Projects, collections, resources, lineage and brand kits browsed and edited, credit balance queried, and a project shared with someone by email (
lib_list,lib_manage,lib_share)
Requirements
An MCP client pointed at https://mcp.soundside.ai/mcp. Two ways to pay: an API key in an Authorization: Bearer header, spending credits at 100 credits to the dollar; or x402 in USDC on Base mainnet, which needs no account. Provider generation is passed through at wholesale rounded up to the cent, platform tools are fixed at $0.01 a call and vision QA at $0.03. Every call is quoted before it runs and the quote is a ceiling, never a floor. Prices are database-driven — read https://mcp.soundside.ai/api/x402/status rather than hardcoding them. As of 2026-08-23 Luma is gone entirely and Runway is audio-only.
Setup effort
Add the hosted endpoint plus a key — paste it into your client, then supply credentials
USE CASES
01
Turning a brief into a finished video
Hand over the plan and let the composition pipeline enrich it, generate the assets in parallel and assemble them with transitions, audio ducking and overlays. One call instead of a dozen, and the parallel generation is where the time goes.
02
Keeping a character consistent across shots
Generate the first image, then pass its resource id as the character reference on every following image or video call. Consistency is the thing that usually breaks when shots are generated independently.
03
Cutting footage you already have
Trim, join, crossfade, retime and colour-grade material in the library, add text or a split screen, and pull out the frames you need. Raw FFmpeg is there for the transform nothing else covers.
04
Checking a render before it ships
Run analysis over the output for technical metadata, a vision pass on what is actually on screen, a transcript, and segment boundaries. Cheaper than the review meeting that finds the same problems.
05
Producing the deck as well as the film
Charts, slides, documents and diagrams come from the same endpoint, and bundle mode returns several formats from one brief — the deck and the chart in a single call rather than two tools and a hand-off.
06
Paying without opening an account
Use the x402 path: USDC on Base, signed off-chain per call, with the facilitator paying gas. Read the live status endpoint first — the published amount is a worst-case ceiling and the settled charge is usually lower.
07
Teaching it your own look
Train a LoRA adapter from media already in your library, deploy it, and generate against it. When the look changes, point the adapter at a different checkpoint rather than retraining from scratch.
19 tools
Listed in project docs. Your client may show a different list.
create_videocreate_audiocreate_musiccreate_textcreate_artifactcompose_videoedit_videoedit_audiocompose_mediaapply_effectextract_mediaanalyze_mediatrain_adapterlist_adaptersmanage_adapterlib_listlib_managelib_shareBEFORE YOU INSTALL
What to know first
Tools you get
19 tools named in the project's documentation.
Where it runs
The documented client connects to a hosted endpoint.
INSTALLATION
Connect your client
Server setup
ENDPOINT URL
https://mcp.soundside.ai/mcpAdd this hosted endpoint to your MCP client using the credentials listed under Requirements.
MCP Clients
Any MCP client — HTTP
ANY MCP CLIENT — HTTP · JSON
{
"mcpServers": {
"soundside-docs": {
"url": "https://mcp.soundside.ai/mcp",
"headers": {
"Authorization": "Bearer <YOUR_API_KEY>"
}
}
}
}
