io.github.mordor-forge/gemini-media-mcp

What it mainly works on, as our AI model read its tool descriptions: files and code on the machine it runs on.

9 of 14 tools are described as changing something (creating, editing, deleting, sending or running).

• edit_imageon 26 other serverscreates or changes datadescribed as changing data

  “Change an existing image with an instruction (add/remove/replace elements, restyle, relight, change aspect ratio, fix text) while keeping everything else.”

  Publisher declares: read-only no · destructive no · idempotent no · open-world yes

  8 inputs, 2 required

  • imagestringrequired“The image to edit: file path, gemini-media:// URI from an earlier result, or data: URI.”
  • promptstringrequired“The change to make, stated specifically, plus what must stay the same (e.g.”
  • referenceImagesnull or array“Optional extra images to pull elements or style from (e.g.”
  • modelstring“Model ID or alias.”
  • aspectRatiostring“Change the aspect ratio (outpainting).”
  • imageSizestring“Output size: 1K, 2K or 4K (512 only on the deprecated gemini-3.1-flash-image).”
  • outputNamestring“Optional base file name for the saved file.”
  • approvedCostUsdnumber“Only needed above the confirmation threshold: the cost in USD the user approved.”
• edit_videoon 9 other serverscreates or changes datadescribed as changing data

  “Change an existing clip with an instruction (add, remove or replace objects, restyle, relight, change the background) with Gemini Omni Flash, keeping everything else.”

  Publisher declares: read-only no · destructive no · idempotent no · open-world yes

  9 inputs, 1 required

  • jobIdstring“A completed video job to edit.”
  • videostring“Or a video file to edit: file path, gemini-media:// URI or data: URI.”
  • promptstringrequired“The change, stated simply, ending with 'Keep everything else the same.' (e.g.”
  • referenceImagesnull or array“Optional images to bring into the clip; refer to them as <IMAGE_REF_0>, <IMAGE_REF_1>...”
  • modelstring“Model ID or alias; default omni (the only model that edits videos).”
  • resolutionstring“360p, 720p (default), 1080p or 4k (1080p and 4k are upscaled).”
  • waitSecondsinteger“Block up to this many seconds for the result (default 0 = return a job handle immediately).”
  • outputNamestring“Optional base file name for the saved video.”
  • approvedCostUsdnumber“Only needed above the confirmation threshold: the cost in USD the user approved.”
• estimate_coston 48 other serversreads or searches

  “Estimate the cost of an image, video, speech or music request before running it, optionally comparing all models; also reports remaining budget and whether approval is needed.”

  Publisher declares: read-only yes · idempotent yes · open-world no

  10 inputs, 1 required

  • mediaTypestringrequired“image, video, speech or music”
  • modelstring“Model ID or alias (default model when omitted)”
  • countinteger“Number of images/clips/songs (default 1)”
  • imageSizestring“Images: 512, 1K, 2K or 4K”
  • inputImagesinteger“Images: number of reference/source images”
  • resolutionstring“Video: 720p, 1080p or 4k (Omni also 360p)”
  • durationSecondsinteger“Video: clip length (default 8)”
  • generateAudionull or boolean“Video on Vertex AI: false for the cheaper silent rate”
  • textstring“Speech: the text to be spoken (length drives cost)”
  • compareboolean“Also estimate the same request on every other active model of this media type”
• extend_videoon 7 other serverscreates or changes datadescribed as changing data

  “Continue a completed clip from its end: omni adds up to 10 s (40 s total, keeping characters and audio coherent); Veo fast/standard add about 7 s (720p, up to 148 s, within 2 days).”

  Publisher declares: read-only no · destructive no · idempotent no · open-world yes

  7 inputs, 2 required

  • jobIdstringrequired“Job handle of a completed generate_video or extend_video call (operation names from older versions also work).”
  • promptstringrequired“What happens next.”
  • negativePromptstring
  • seednull or integer“Vertex AI only.”
  • waitSecondsinteger“Block up to this many seconds for the result (default 0).”
  • outputNamestring
  • approvedCostUsdnumber“Only needed above the confirmation threshold.”
• generate_imageon 44 other serversdescribed as changing data

  “Create images from a text prompt, optionally guided by up to 14 reference images (subjects, products, style).”

  Publisher declares: read-only no · destructive no · idempotent no · open-world yes

  9 inputs, 1 required

  • promptstringrequired“What to create.”
  • modelstring“Model ID or alias.”
  • aspectRatiostring“e.g.”
  • imageSizestring“Output size: 1K (default), 2K or 4K.”
  • referenceImagesnull or array“Optional images to guide the result (subjects, products, characters, style): file paths, gemini-media:// URIs from earlier results, or data: URIs.”
  • countinteger“Number of variations to generate in parallel (1-4, default 1).”
  • googleSearchboolean“Ground the image in Google Search results (current events, real places, data for infographics).”
  • outputNamestring“Optional base file name (without extension) for the saved file(s).”
  • approvedCostUsdnumber“Only needed when a call exceeds the configured confirmation threshold: the cost in USD the user approved.”
• generate_musicon 10 other serversgenerates contentdescribed as changing data

  “Compose music with Lyria: 30-second clips (clip, $0.04) or full songs of a few minutes with vocals and lyrics (full, $0.08).”

  Publisher declares: read-only no · destructive no · idempotent no · open-world yes

  10 inputs, 1 required

  • promptstringrequired“Describe the music: genre, mood, instruments, tempo, key, production style and song structure.”
  • modelstring“clip (default: 30-second MP3, cheapest) or full (complete song up to about 3 minutes, vocals, WAV option).”
  • lyricsstring“Optional lyrics with section tags; they are sung in the language they are written in.”
  • instrumentalboolean“Ask for no vocals.”
  • bpminteger“Optional tempo hint in beats per minute.”
  • durationSecondsinteger“Optional target length hint for full songs (clip is always 30s).”
  • imagesnull or array“Optional images (up to 10) whose mood or scene should inspire the music.”
  • formatstring“mp3 (default) or wav (full model only).”
  • outputNamestring“Optional base file name.”
  • approvedCostUsdnumber“Only needed above the confirmation threshold.”
• generate_speechon 2 other serversgenerates contentdescribed as changing data

  “Turn text into natural speech (WAV) with one voice, or a dialogue between two speakers.”

  Publisher declares: read-only no · destructive no · idempotent no · open-world yes

  10 inputs

  • textstring“The exact words to speak (single speaker).”
  • dialoguenull or array“Alternative to text: a conversation between up to 2 named speakers, one entry per line.”
  • voicestring“Voice for single-speaker text, e.g.”
  • speakersnull or array“Voice per dialogue speaker; unlisted speakers get distinct defaults.”
  • stylestring“How to deliver the lines, e.g.”
  • languageCodestring“Optional BCP-47 code (e.g.”
  • modelstring“Model ID or alias: tts (default, Gemini 3.8 Flash TTS), tts-lite (cheaper).”
  • formatstring“wav (default) or pcm (raw 16-bit little-endian).”
  • outputNamestring“Optional base file name.”
  • approvedCostUsdnumber“Only needed above the confirmation threshold.”
• generate_videoon 21 other serversgenerates contentdescribed as changing data

  “Start a video clip with native audio from a text prompt, optionally from a first frame, first+last frames, or reference images.”

  Publisher declares: read-only no · destructive no · idempotent no · open-world yes

  15 inputs, 1 required

  • promptstringrequired“What happens in the clip: subject and action, camera movement, setting, lighting, style, and sound (dialogue in quotes, sound effects, ambience, music).”
  • modelstring“Model ID or alias: omni (default on the Gemini API: Gemini Omni Flash, best prompt adherence, editable and extendable to 40 s), lite (cheapest Veo; the default…”
  • aspectRatiostring“16:9 (default) or 9:16”
  • resolutionstring“720p (default), 1080p or 4k (not lite); omni also 360p for fast drafts.”
  • durationSecondsinteger“Clip length: Veo 4, 6 or 8 seconds; omni 3-10 seconds (default 8).”
  • imagestring“Optional first frame (image-to-video): file path, gemini-media:// URI or data: URI.”
  • lastFramestring“Optional last frame; requires image.”
  • referenceImagesnull or array“Images of characters, objects, products or styles to keep consistent: Veo fast/standard up to 3 (forces 8s, not with image/lastFrame); omni up to 10 including …”
  • negativePromptstring“What to avoid, as plain nouns (e.g.”
  • seednull or integer“Seed for more repeatable results (Vertex AI only; ignored on the Gemini API).”
  • personGenerationstring“allow_all, allow_adult or dont_allow.”
  • generateAudionull or boolean“Vertex AI only: set false for silent video at a lower price.”
  • waitSecondsinteger“Block up to this many seconds for the result (default 0 = return a job handle immediately).”
  • outputNamestring“Optional base file name for the saved video.”
  • approvedCostUsdnumber“Only needed above the confirmation threshold: the cost in USD the user approved.”
• get_configon 21 other serversreads or searches

  “Show the active backend and how it was chosen, output directory, default models, budgets, input-file policy and configuration warnings.”

  Publisher declares: read-only yes · idempotent yes · open-world no

  Takes no inputs.

• get_usageon 99 other serversreads or searches

  “Report estimated spend for this session, today, this month or all time, broken down by model and tool, with budget remaining, recent calls and running video jobs.”

  Publisher declares: read-only yes · idempotent yes · open-world no

  2 inputs

  • periodstring“session (since this server started), today (default), month or all”
  • recentinteger“How many recent entries to include (default 10, max 100)”
• get_videoon 17 other serversreads or searches

  “Check a video job and wait for it (default up to 45s).”

  Publisher declares: read-only no · destructive no · idempotent yes · open-world yes

  2 inputs, 1 required

  • jobIdstringrequired“Job handle returned by generate_video, extend_video or edit_video.”
  • waitSecondsnull or integer“Block up to this many seconds for completion (default 45, 0 = check once).”
• list_modelson 94 other serversreads or searches

  “List current models per media type with aliases, status (GA/preview/deprecated), prices and a short guide to when to use each.”

  Publisher declares: read-only yes · idempotent yes · open-world yes

  4 inputs

  • mediaTypestring“Filter: image, video, speech or music.”
  • includeInactiveboolean“Also list retired models and their replacements.”
  • liveboolean“Also query the Gemini API for the models this key can use, flagging unavailable ones and media models missing from the catalog.”
  • detailboolean“Include full capabilities (aspect ratios, voices, rules).”

Each quoted sentence is the first sentence of the publisher’s own description. The kind of tool and the “changes data” mark are our AI model’s reading of that description, not something we tested: two differently worded runs had to agree, otherwise no label is shown, and nobody has checked them by hand. A tool without the “changes data” mark is described as only reading, searching or calculating, or the two runs did not agree. No tool was called. Read the publisher’s full description before relying on it. “Publisher declares” lists the hints the server itself sends with the tool (MCP annotations: read-only, destructive, idempotent, open-world); they are the publisher's claims, not checked by us. Inputs are the parameters in the tool's own input schema: name, declared type, whether it is required, and the first sentence of its description. Read 2026-10-23.

The registry verifies a package once, when it is published, and does not re-check it. A difference below is about the listing, not a judgement of the publisher.