MCP App Store
by KeenEthics
Freebeat Music Video Generator icon

Freebeat Music Video Generator MCP Server for ChatGPT

By Freebeat

Overview

Generate music videos, dance videos, and animated lyric videos from your songs. Attach audio in ChatGPT or share a link to it, and add photos to feature people or characters. Freebeat plans, generates, and edits the video. Sign in with your Freebeat account. Generations use your existing credits and usually take 15–60 minutes. By default videos use the first 15 seconds of the song. Let the director decide or co-create through the assistant, with progress and media visible in the conversation. Also create original music, images, short video clips, singing photos, and lip-sync videos.

How to connect

  1. ChatGPT: Open the Freebeat Music Video Generator listing in the ChatGPT app directory and choose Connect. ChatGPT listing Freebeat Music Video Generator exposes 24 tools here.

Tools

edit_image

ChatGPT
Spends the user's Freebeat credits. quote_generation returns the current price or reports that the total is unknown. The server submits only with max_credits at or above the current quote, or with accept_unknown_cost:true when no total can be quoted. Agents charge per stage; no whole-run spending cap is enforced. Use this when the user wants to edit one existing image with a text instruction. Upload the image with upload_image first. Flow: list_models(type=image_edit) -> get_model_help -> edit_image -> get_task_status with the returned batch_id. For a new image without a source, use generate_image. For song-driven music, dance, or lyric videos, use generate_music_video, generate_dance_video, or generate_lyric_video.

generate_dance_video

ChatGPT
Spends the user's Freebeat credits. quote_generation returns the current price or reports that the total is unknown. The server submits only with max_credits at or above the current quote, or with accept_unknown_cost:true when no total can be quoted. Agents charge per stage; no whole-run spending cap is enforced. Freebeat videos are made by a director agent that works in stages, as on freebeat.ai, where a person chats with it and sees what it made. In auto mode it makes the creative decisions and delivers the film; otherwise it pauses for the person's input where taste matters. get_video_session shows its words, stage statuses and work; send_video_agent_message passes on the person's feedback, approval or direction, or hands the rest to auto. Work that follows a message is charged per stage like the rest of the session. A song whose audio is still being analyzed returns preparing with its numeric music_id and retry_after_seconds instead of a task_id; no paid video starts before the audio is ready. Use this for dance video generation: the user wants dancers performing to a song. Freebeat's dance video agent plans and generates a dance video timed to the song; people in optional photos become the dancers. Supports vertical 9:16 and 3:4 formats. A 2026-10-27 production diagnosis found 855 credits for the video step alone of a 15-second 720p dance, with planning and images extra: budget on the order of 1,000 credits, not a fixed quote or cap. Requires music_id from upload_audio or a completed generate_music result; optional photos via upload_image. Starts a generation job that uses the user's existing Freebeat credits and usually takes 15–60 minutes; by default it uses the first 15 seconds of the song, or up to the song's end if shorter. Returns task_id and session_url; poll get_task_status every few minutes, call get_task_result after completed; on action_required read get_video_session and use send_video_agent_message to continue. Do not use for story-based music videos, lyric videos, short singing-photo clips, lip-sync of an existing clip, standalone video clips without a song, or generating audio.

generate_effect

ChatGPT
Spends the user's Freebeat credits. quote_generation returns the current price or reports that the total is unknown. The server submits only with max_credits at or above the current quote, or with accept_unknown_cost:true when no total can be quoted. Agents charge per stage; no whole-run spending cap is enforced. Use this when the user wants an effect-template video from a music_id and exactly one reference image URL. Flow: list_effects -> optional upload_audio -> optional upload_image -> generate_effect -> get_task_status -> get_task_result. By default, use list_effects.image_url. You can also use a single uploaded image URL from upload_image. Call get_task_result after get_task_status returns completed, failed, or action_required; failure and billing remain available without final media. For song-driven music, dance, or lyric videos, use generate_music_video, generate_dance_video, or generate_lyric_video.

generate_image

ChatGPT
Spends the user's Freebeat credits. quote_generation returns the current price or reports that the total is unknown. The server submits only with max_credits at or above the current quote, or with accept_unknown_cost:true when no total can be quoted. Agents charge per stage; no whole-run spending cap is enforced. Use this when the user wants new still images from a text prompt. Flow: list_models(type=image) -> get_model_help -> generate_image -> get_task_status with the returned batch_id. For changes to an existing image, use edit_image. For song-driven music, dance, or lyric videos, use generate_music_video, generate_dance_video, or generate_lyric_video.

generate_lip_sync_video

ChatGPT
Spends the user's Freebeat credits. quote_generation returns the current price or reports that the total is unknown. The server submits only with max_credits at or above the current quote, or with accept_unknown_cost:true when no total can be quoted. Agents charge per stage; no whole-run spending cap is enforced. Use this when the user wants to lip-sync an existing video clip to a track. Upload the clip with upload_video and audio with upload_audio first, then poll get_task_status and fetch get_task_result. For a singing photo, use generate_singing_video. For song-driven music, dance, or lyric videos, use generate_music_video, generate_dance_video, or generate_lyric_video.

generate_lyric_video

ChatGPT
Spends the user's Freebeat credits. quote_generation returns the current price or reports that the total is unknown. The server submits only with max_credits at or above the current quote, or with accept_unknown_cost:true when no total can be quoted. Agents charge per stage; no whole-run spending cap is enforced. Freebeat videos are made by a director agent that works in stages, as on freebeat.ai, where a person chats with it and sees what it made. In auto mode it makes the creative decisions and delivers the film; otherwise it pauses for the person's input where taste matters. get_video_session shows its words, stage statuses and work; send_video_agent_message passes on the person's feedback, approval or direction, or hands the rest to auto. Work that follows a message is charged per stage like the rest of the session. A song whose audio is still being analyzed returns preparing with its numeric music_id and retry_after_seconds instead of a task_id; no paid video starts before the audio is ready. Use this for lyric video generation: the user wants animated lyrics timed to the vocals of a song. Freebeat's lyric video agent plans, generates, and edits the lyric video. Requires music_id from upload_audio or a completed generate_music result; optional reference photos via upload_image. Starts a generation job that uses the user's existing Freebeat credits and usually takes 15–60 minutes; by default it uses the first 15 seconds of the song, or up to the song's end if shorter. Returns task_id and session_url; poll get_task_status every few minutes, call get_task_result after completed; on action_required read get_video_session and use send_video_agent_message to continue. Do not use for story-based music videos, dance videos, short singing-photo clips, lip-sync of an existing clip, standalone video clips without a song, or generating audio.

generate_music

ChatGPT
Spends the user's Freebeat credits. quote_generation returns the current price or reports that the total is unknown. The server submits only with max_credits at or above the current quote, or with accept_unknown_cost:true when no total can be quoted. Agents charge per stage; no whole-run spending cap is enforced. Use this when the user wants to generate original music from a prompt, with optional lyrics. Call list_music_models first to pick a model; only models with a quoted current price can be submitted; then poll get_task_status and fetch get_task_result with the returned task_id. Does not turn an existing song into a video. For song-driven music, dance, or lyric videos, use generate_music_video, generate_dance_video, or generate_lyric_video.

generate_music_video

ChatGPT
Spends the user's Freebeat credits. quote_generation returns the current price or reports that the total is unknown. The server submits only with max_credits at or above the current quote, or with accept_unknown_cost:true when no total can be quoted. Agents charge per stage; no whole-run spending cap is enforced. Freebeat videos are made by a director agent that works in stages, as on freebeat.ai, where a person chats with it and sees what it made. In auto mode it makes the creative decisions and delivers the film; otherwise it pauses for the person's input where taste matters. get_video_session shows its words, stage statuses and work; send_video_agent_message passes on the person's feedback, approval or direction, or hands the rest to auto. Work that follows a message is charged per stage like the rest of the session. A song whose audio is still being analyzed returns preparing with its numeric music_id and retry_after_seconds instead of a task_id; no paid video starts before the audio is ready. Use this for music video generation: the user wants a music video (MV) for a song. Freebeat's music video agent plans scenes and generates and edits a multi-scene video timed to the song; people or characters from optional photos can appear. Requires music_id from upload_audio or a completed generate_music result; optional photos via upload_image. Starts a generation job that uses the user's existing Freebeat credits and usually takes 15–60 minutes; by default it uses the first 15 seconds of the song, or up to the song's end if shorter. Returns task_id and session_url; poll get_task_status every few minutes, call get_task_result after completed; on action_required read get_video_session and use send_video_agent_message to continue. Do not use for dance videos, lyric videos, short singing-photo clips, lip-sync of an existing clip, standalone video clips without a song, or generating audio.

generate_singing_video

ChatGPT
Spends the user's Freebeat credits. quote_generation returns the current price or reports that the total is unknown. The server submits only with max_credits at or above the current quote, or with accept_unknown_cost:true when no total can be quoted. Agents charge per stage; no whole-run spending cap is enforced. Use this when the user wants a photo, pet, or duet to sing a track as a short performance video. Upload images with upload_image and audio with upload_audio first, then poll get_task_status and fetch get_task_result. For lip-sync of an existing clip, use generate_lip_sync_video. For song-driven music, dance, or lyric videos, use generate_music_video, generate_dance_video, or generate_lyric_video.

generate_text

ChatGPT
Spends the user's Freebeat credits. quote_generation returns the current price or reports that the total is unknown. The server submits only with max_credits at or above the current quote, or with accept_unknown_cost:true when no total can be quoted. Agents charge per stage; no whole-run spending cap is enforced. Use this when the user wants a reply from a selected Freebeat text model, discovered with list_models(type=text). Usage is billed to the user's existing Freebeat credits. Returns text, not media. For song-driven music, dance, or lyric videos, use generate_music_video, generate_dance_video, or generate_lyric_video.

generate_video

ChatGPT
Spends the user's Freebeat credits. quote_generation returns the current price or reports that the total is unknown. The server submits only with max_credits at or above the current quote, or with accept_unknown_cost:true when no total can be quoted. Agents charge per stage; no whole-run spending cap is enforced. Use this when the user wants a standalone video clip from text, images, or reference frames. Supports text-to-video, image-to-video, first/last-frame transition, or reference-to-video when the model supports it. Upload images with upload_image first. Flow: list_models(type=video) -> get_model_help -> generate_video -> get_task_status with the returned batch_id. For song-driven music, dance, or lyric videos, use generate_music_video, generate_dance_video, or generate_lyric_video.

get_credit_balance

ChatGPT
Read the signed-in Freebeat account's available credit balance. Does not charge credits or start generation.

get_model_help

ChatGPT
Use this when the user needs supported parameters and credit costs for one image, image_edit, or standalone video model, read live from Freebeat. pricing.keys name the parameters that change the price and pricing.prices contains the rule body; combination tables are in pricing.prices.map, with duration units and multipliers retained. Freebeat validates every request; if a generation is rejected, its message says which parameter to change. This does not start generation or provide text or music model help. For song-driven music, dance, or lyric videos, use generate_music_video, generate_dance_video, or generate_lyric_video.

get_task_result

ChatGPT
Use this when the user needs output or failure and billing details from an existing Freebeat task. Call after get_task_status returns completed, failed, or action_required. Failed or paused tasks return failure and billing even without final media. When an agent video is action_required, read get_video_session and use send_video_agent_message with the same task_id to continue in either mode. Polling a song-driven video may start its final video assembly from already generated assets; it does not regenerate source assets. Song-driven videos from generate_music_video, generate_dance_video, and generate_lyric_video, plus effects, singing, and lip-sync tasks, return result.video_url and usually result.cover_url; music tasks return result.audio_url. Song-driven video assembly still in progress returns running with a message to poll again. Other pending or running tasks return TASK_NOT_COMPLETED. If the task failed, this tool surfaces the backend failure. For Open RPC image/video batches, get result URLs from get_task_status instead.

get_task_status

ChatGPT
Use this when the user needs the status of an existing asynchronous Freebeat task. Supports task IDs from generate_music_video, generate_dance_video, generate_lyric_video, music, singing, lip-sync, and effect tools, plus Open RPC image/video batch IDs. Polling a song-driven video may start its final video assembly from already generated assets; it does not regenerate source assets. Song-driven video jobs report running, completed, failed, or action_required; while running, relay the latest step named in message; completed videos include video_url, and on action_required read get_video_session with this task_id and use send_video_agent_message to continue the conversation in either auto or co-create mode. For Open RPC batches, result URLs are returned on completed items; do not call get_task_result for these batches. For other tasks call get_task_result after completed, failed, or action_required. On failures or pauses, explain failure.stage, failure.reason, failure.retryable and billing charged/refunded/net/refund_state; null means not reported, and pending refunds are not refunded credits.

get_video_session

ChatGPT
Freebeat videos are made by a director agent that works in stages, as on freebeat.ai, where a person chats with it and sees what it made. In auto mode it makes the creative decisions and delivers the film; otherwise it pauses for the person's input where taste matters. get_video_session shows its words, stage statuses and work; send_video_agent_message passes on the person's feedback, approval or direction, or hands the rest to auto. Work that follows a message is charged per stage like the rest of the session. Read an existing song-driven video session: recent conversation and requests, running or waiting state, pipeline statuses, auto setting, and readable artifacts with viewable or playable media URLs. fields contain labeled display values, media contains their URLs: plain web links that open in a browser. ChatGPT drops inline images from these hosts, so a person sees them only as links. Results are bounded; truncated means some content was shortened or omitted. This read does not start work or final rendering.

list_effects

ChatGPT
Use this when the user wants effect templates and the metadata needed to select an effect and create an effect task. Returns 20 effects per page by default; use page and page_size to browse, and next_page until it is null. Each effect includes a default image_url and default_music_id you can use directly. This lists templates rather than generating a video. For song-driven music, dance, or lyric videos, use generate_music_video, generate_dance_video, or generate_lyric_video.

list_models

ChatGPT
Use this when the user needs available models for standalone image, image editing, video clip, or text generation. type=image for generate_image, image_edit for edit_image, video for generate_video, text for generate_text. Music and voice models are listed by list_music_models. Then call get_model_help for parameters and prices of one image, image_edit, or video model. This lists models rather than starting generation. For song-driven music, dance, or lyric videos, use generate_music_video, generate_dance_video, or generate_lyric_video.

list_music_models

ChatGPT
Use this when the user needs music-generation models and their supported inputs before calling generate_music. This lists models rather than generating audio or video. Each model's pricing shows its current price; generate_music submits only models whose pricing status is quoted. For song-driven music, dance, or lyric videos, use generate_music_video, generate_dance_video, or generate_lyric_video.

quote_generation

ChatGPT
quote_generation returns the current price or reports that the total is unknown. The server submits only with max_credits at or above the current quote, or with accept_unknown_cost:true when no total can be quoted. Agents charge per stage; no whole-run spending cap is enforced. Supply the intended tool and its planned input parameters. Reads current Freebeat pricing rules, without starting or charging for generation. Quoted credits are list price before account discounts or allowances. Unavailable means unknown, never free; agent totals depend on stages and retries. For music, check list_music_models for length control before quoting.

send_video_agent_message

ChatGPT
Passes the person's feedback, approval or direction to their existing Freebeat video session, as the chat box on freebeat.ai does, and can change the session's auto setting. A message is accepted only while the director agent is not running. Work the agent does after a message uses the user's Freebeat credits. The agent's reply appears in get_video_session.

upload_audio

ChatGPT
Use this when the user needs to upload an attached audio file, upload a local audio file in stdio mode, or import from a link: a public HTTPS link to an audio file, or a music page on SoundCloud, YouTube, Suno, Udio, TikTok, Stable Audio, or Riffusion. Provide exactly one of file, file_path, or url. Returns music_id for generate_music_video, generate_dance_video, generate_lyric_video and other audio workflows. Does not generate audio or video.

upload_image

ChatGPT
Use this when the user needs to upload one or two image files (PNG/JPG/JPEG, max 10MB each) for Freebeat workflows. Use returned image_urls with generate_music_video, generate_dance_video, or generate_lyric_video for song-driven videos, or pass a single uploaded image URL to generate_effect.reference_image_urls to replace the default list_effects.image_url. Does not generate or edit images.

upload_video

ChatGPT
Use this when the user needs to upload an MP4, MOV, or WebM video for a Freebeat lip-sync workflow from an attachment or public HTTPS url. Returns a video_url for generate_lip_sync_video. Does not generate a new clip. For song-driven music, dance, or lyric videos, use generate_music_video, generate_dance_video, or generate_lyric_video.

Capabilities

Writes

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/freebeat-music-video-generator"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/freebeat-music-video-generator)