MCP App Store
by KeenEthics
ProsodyAI icon

ProsodyAI MCP Server for ChatGPT

By ProsodyAI

Overview

ProsodyAI turns text into speech and descriptions into video, on your own ProsodyAI account. Speech: 59 languages with OpenAI or Gemini voices. Describe the delivery in plain words, such as "warm and unhurried" or "excited sports commentator", and get a download link. Video: describe a scene and ProsodyAI renders a clip with Veo, Seedance or Wan 3.0. A clip takes a few minutes; ask for its status and you get the link when it is ready. Characters: create a character from a description, approve its front view, and reuse it so the same person appears across clips. Edits: add a voiceover to a video ProsodyAI made, with optional burned-in subtitles; merge clips, trim, remove silences or make a GIF. Account: see your plan and what is left this month. Who it is for: creators, marketers and teams who need narration, voiceovers and short clips without leaving the conversation. Limitations: you sign in with a ProsodyAI account, and every request uses that account's plan and its allowance; video needs video seconds on the account. Nothing is sold inside ChatGPT. There is no voice cloning: speech uses ProsodyAI's built-in voices. Edits work only on videos ProsodyAI made for you, not on files uploaded to the chat. Download links stay valid for 7 days, and results also appear in your history at prosodyai.ai/history.

How to connect

  1. ChatGPT: Open the ProsodyAI listing in the ChatGPT app directory and choose Connect. ChatGPT listing ProsodyAI exposes 16 tools here.

Tools

approve_character

ChatGPT
Approve a character's front view and draw its remaining views (images from the plan's allowance).

burn_captions

ChatGPT
Burn the caption lines into the video they were made from, as a new video. Uses the lines as they are now, so make corrections with edit_captions first. Free: it spends no video seconds or caption minutes. style: classic (white text at the bottom, the default) or bold_center. Returns a job id for get_job (kind video).

create_character

ChatGPT
Create a reusable character from a description; ProsodyAI draws a front view (one image from the plan's image allowance). Approve it with approve_character to draw the other views, then use its id in generate_video for the same person across clips.

edit_captions

ChatGPT
Correct the text of caption lines made by transcribe_video. Send only the lines that change, by their index from get_job (kind captions); times and the number of lines stay as they are. Each line holds 1 to 200 characters. Free.

edit_video

ChatGPT
Edit videos ProsodyAI made for this user, in the Studio. Edits are rendered from existing files and do not draw on video seconds or speech characters (a voiceover uses speech already made with generate_speech). Operations: add_voiceover (video_job_id + speech_id from generate_speech, optional burned-in subtitles), merge (two or more video_job_ids), trim (video_job_id, start, end in seconds), remove_silences, make_gif. Returns a job id for get_job.

generate_speech

ChatGPT
Turn text into speech on the user's ProsodyAI account. Spends one premium character per character of text from the plan's monthly allowance. style is how it should sound, in plain words ("warm, unhurried"). Voice choice: give voice_id from list_voices and its own provider is used; otherwise give provider (openai_tts or google_tts) and the first voice of that provider for the language is used; with neither, OpenAI (openai_tts) is tried first, then Gemini (google_tts). Returns a download link, or a task to check with get_job.

generate_video

ChatGPT
Generate a video clip from a description on the user's account. Spends the plan's video seconds (see list_video_options for what each engine and length costs). Takes a few minutes: returns a job id to check with get_job. model is the veo engine's resolution (720p, 1080p, 4k — veo is Gemini Omni), the Seedance model id, or the Wan resolution (480p, 720p, 1080p). seconds: veo 4, 6, 8, 10 or 15, Seedance 4-30 (the model snaps it to a length it offers; 2.0 up to 15), Wan 2-30.

get_account

ChatGPT
The user's plan and what is left this month: premium characters, video seconds, images and sound effects. Free.

get_job

ChatGPT
Check a speech task (kind speech, the task id), a video job (kind video, the job id) or a sound effect (kind sound_effect, the sound id) and get its link when finished. Free.

list_characters

ChatGPT
The user's characters with their id and stage. Free.

list_templates

ChatGPT
Ready-made video looks (Studio templates): product spins, logo reveals, app demos, property tours, food close-ups, seasonal promos and more. Each has fields to fill; most need one photo, which make_template_video takes as an attachment or a link. Listing costs nothing.

list_video_options

ChatGPT
What video the user's plan can make: engines (veo = Gemini Omni, seedance, wan), models, lengths, resolutions and what each costs. Free.

list_voices

ChatGPT
List ProsodyAI's voices, optionally for one language (code such as en, tr, de) or one provider (openai_tts, google_tts), and the style presets. Free.

make_sound_effect

ChatGPT
Make one sound effect (a mono WAV) from a synthesis recipe you write: stylised, synthesised sounds for games, videos and apps (arcade, UI, whooshes, impacts, stingers), royalty-free. Not realistic recordings, not speech, not music tracks. One sound per call: call it again for each sound of a set. To change a sound, edit the recipe and call again. Paid plans only; each call spends one of the plan's monthly sound effects. Recipe (JSON object): { "duration": 0.6, // seconds, whole sound including tails, max 12 "seed": 7, // optional, fixes the noise "layers": [ LAYER, ... ], // 1..24, mixed together "master": { // optional "delay": {"time": 0.12, "feedback": 0.3, "mix": 0.25}, "reverb": {"mix": 0.2, "decay": 1.2, "size": 1.0, "damping_hz": 6000}, "peak": 0.9 // output is normalised to this peak } } LAYER: { "source": {"type": "square"}, // sine | square | saw | triangle | pulse | fm // | noise | pink_noise | brown_noise "start": 0.0, // seconds into the sound "duration": 0.3, // seconds (required) "gain": 0.8, "pitch": CURVE, // Hz, default 440 "vibrato": {"rate": 6, "depth_semitones": 0.3}, "amp_env": CURVE, // 0..1 over the layer's time, default 1 "filters": [{"type": "lowpass" | "highpass" | "bandpass", "cutoff": CURVE, "q": 0.707}], // max 4 "drive": 0.0, // 0..1 tanh saturation "bitcrush": {"bits": 6, "downsample": 4}, "repeat": {"count": 4, "interval": 0.06, "gain_decay": 0.8, "pitch_step_semitones": 4} } Source extras: pulse "duty": CURVE (0.01..0.99); fm "mod_ratio": 1.4 and "mod_index": CURVE (brightness, bells about 3 -> 0); any noise "hold_pitch": true turns pitch into a sample-and-hold rate for crunchy 8-bit noise (without it, pitch is ignored for noise: shape noise with filters). CURVE: a number, a list of [time_seconds, value] points (linear, time from the layer start, max 64 points), or {"points": [[t, v], ...], "interp": "exp"} for exponential sweeps (pitch and cutoff; values above 0). Limits: unknown fields are refused; layer duration x repeat count summed over layers must stay under 120 s. Tips: envelopes 0 -> peak in 1-5 ms -> 0; impacts = noise transient + low body with a pitch drop + tail; retro = square/pulse + exp pitch sweeps + bitcrush; whoosh = noise through a sweeping bandpass; melodies = one layer per note.

make_template_video

ChatGPT
Make a video from a ready-made look (see list_templates) on the user's account, e.g. a product spin or a logo reveal from the user's photo. Spends the plan's video seconds like generate_video (on a one-time pack, its characters); the template sets the length, model and format. Pass the template's text, choice and character fields in fields. For the photo, use the image the user attached in the chat (photo) or a public link to it (photo_url). Paid plans, or a one-time pack. Takes a few minutes: returns a job id to check with get_job (kind video).

transcribe_video

ChatGPT
Make captions (subtitles) for a video ProsodyAI made for this user: the speech in it is written out as timed lines. Paid plans only; the video's length, rounded up to whole minutes, is spent from the month's caption minutes (see get_account). Returns a captions id: check it with get_job (kind captions) to read the lines, fix words with edit_captions, then burn them in with burn_captions. language is the spoken language; leave it out to detect it. Only ProsodyAI videos can be captioned from the chat; for other files the user can use https://prosodyai.ai/captions.

Capabilities

Writes

Example Prompts

Click any prompt to copy it.

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/prosodyai"><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/prosodyai)