Skip to main content
Glama

Server Details

Generate images, video and audio from 120+ AI models, plus lipsync, headshot and cast-swap studios.

Ownership verified
Status
Healthy
Uptime
8.7% over the last 24 hours
OAuth
Works in Glama
Last Tested
Transport
Streamable HTTP · MCP 2025-11-25
URL
ChatGPT
–
Claude
–

TDQS

B3.2/5.0

Scored across 28 tools

Disambiguation4/5

Most tools have clearly distinct purposes: generation lifecycle (generate/get/cancel/list/wait), catalogue lookups (list_models, get_model, list_voices, list_headshot_styles, list_product_templates), and media library operations are all separable. The riskiest cluster is upload_file vs request_upload vs get_upload, but the descriptions explicitly disambiguate (public URL vs user-device upload vs retrieval), and the workflow wrappers (headshot, product_shot, try_on, cast_swap) are framed as curated pipelines over `generate`. Only mild overlap remains between `generate` and those wrappers.

Naming Consistency4/5

Two consistent families: snake_case verb_noun for infrastructure operations (get_generation, list_models, cancel_generation, request_upload, switch_workspace) and noun/snake_case names for the content-producing workflows (headshot, try_on, lipsync, product_shot, text_to_speech). The verb-prefixed tools are predictable and the noun group is internally coherent, though the two conventions side by side are a minor deviation from a single pattern.

Tool Count3/5

At 28 tools the surface is heavy, but the domain genuinely spans image, video, audio, uploads, billing and workspace management, so most tools map to a real capability rather than filler. The curated wrappers (headshot, product_shot, try_on, cast_swap, lipsync, talking_head_*) could in principle be folded into `generate` plus a model id, which makes the set feel larger than strictly necessary.

Completeness4/5

Coverage is strong across the whole lifecycle: quoting (estimate_cost), starting, polling, cancelling, listing and inspecting generations, plus upload intake, file listing, model/style/voice/template catalogues, account balance and workspace switching. Remaining gaps are minor: no delete/cleanup for files or generations, no output-download/export helper, and no way to update or re-run a job from an existing one.

Available Tools

28 tools
cancel_generationCancel generationB
Destructive
Inspect

Cancel a queued generation (refunded). A render already running at the provider may not be cancellable.

Inputs: generation_id (required). Returns: object

cast_swapCast swapBInspect

Replace the people in a 4–30 s clip with the people in your photos. subjects: [{photos: [file ids], kind?: person|object, label?, slot?: left|center|right|main, placement?}], up to 4 subjects / 9 photos.

Inputs: subjects (required), source_video_file (required), notes, resolution, +2 more. Returns: object

enhance_promptEnhance promptB
Read-only
Inspect

Rewrite a prompt the way the studio's prompt improver does. kind: image | video.

Inputs: prompt (required), kind, model_id. Returns: object

estimate_costEstimate costA
Read-only
Inspect

Validate an input and quote the credits a generation would charge, without charging. The same validation generate runs.

Inputs: input (required), model_id (required). Returns: object

generateGenerateAInspect

Start a generation (charges credits up front; refunded if it fails or is cancelled). Returns the job; poll with wait_for_generation. Above the workspace's confirm threshold the call returns status: confirmation_required with the quoted credits — ask the user, then call again with confirm_credits equal to that quote. wait_seconds (≤ 25) blocks for quick models.

Inputs: input (required), model_id (required), wait_seconds, confirm_credits. Returns: object

get_accountGet accountA
Read-only
Inspect

The connected workspace, its credit balance, today's agent spend and the agent limits.

Inputs: no named parameters. Returns: object

get_generationGet generationC
Read-only
Inspect

Current status and outputs of a generation.

Inputs: generation_id (required). Returns: object

get_modelGet modelC
Read-only
Inspect

Parameters (with enums, ranges, defaults and whether they change the price), input slots for files, capabilities, credit pricing and an example request for one model.

Inputs: model_id (required). Returns: object

get_uploadGet uploadA
Read-only
Inspect

The files the user added through a request_upload link. status is ready (use files[].id in a model's input slots or a studio tool), waiting (nothing uploaded yet: ask the user, do not poll in a loop) or expired (call request_upload again).

Inputs: request_id (required). Returns: object

headshotHeadshotCInspect

A professional headshot of the person in face_file (or a saved character) in a curated style and/or with explicit controls (list_headshot_styles).

Inputs: notes, aspect, controls, face_file, +3 more. Returns: object

lipsyncLipsyncBInspect

Sync a video's mouth to an audio track (file ids from upload_file or previous outputs). Price depends on the rendered length; the result's credits_charged is the quote.

Inputs: audio_file (required), video_file (required), model, length_mode, +1 more. Returns: object

list_charactersList charactersA
Read-only
Inspect

The user's saved characters (for headshot and image models taking character_id).

Inputs: no named parameters. Returns: object

list_filesList filesC
Read-only
Inspect

Files in the workspace library: uploads and generation outputs.

Inputs: type, limit. Returns: object

list_generationsList generationsB
Read-only
Inspect

Recent generations made through this connection's workspace by agents, newest first. status: queued | processing | succeeded | failed | cancelled.

Inputs: limit, status. Returns: object

list_headshot_stylesList headshot stylesB
Read-only
Inspect

Curated headshot styles (use id as style_item_id) and the control vocabulary.

Inputs: no named parameters. Returns: object

list_modelsList modelsA
Read-only
Inspect

List the models you can generate with. Filter by feature slug (e.g. image-generation, image-editing, text-to-video, image-to-video, lipsync, text-to-speech), output_type (image | video | audio) or a query substring. Prices are credits.

Inputs: query, feature, output_type. Returns: object

list_product_templatesList product templatesB
Read-only
Inspect

Scene templates for product_shot (use key).

Inputs: no named parameters. Returns: object

list_talking_head_setsList talking-head setsB
Read-only
Inspect

Curated background sets for talking_head_keyframe.

Inputs: no named parameters. Returns: object

list_voicesList voicesA
Read-only
Inspect

Library voices for text_to_speech (voice value, name, language, gender, models).

Inputs: no named parameters. Returns: object

product_shotProduct shotCInspect

An ad photo of the uploaded product in a scene template (list_product_templates); brief adds copy or direction.

Inputs: product_file (required), template_key (required), brief, aspect, +2 more. Returns: object

request_uploadRequest uploadAInspect

Get a link where the user can upload files from their device (a photo of themselves, a product image, a clip, a voice track).

    Use this whenever the user refers to a file you cannot give a
    public URL for, including anything they attached to this chat:
    attachments cannot be passed to Affogato. Show the user
    `upload_url` as a clickable link and ask them to tell you when they
    are done, then call `get_upload` with `request_id` for the file ids.
    `purpose` is shown on the page: say what to upload ("A clear photo
    of your face"). `kind` is image, video, audio or any. The link
    works for one hour and only for a signed-in member of this
    workspace. Costs no credits.

Inputs: purpose (required), kind, max_files. Returns: object

switch_workspaceSwitch workspaceAInspect

Bind this connection to another workspace (team) the user belongs to. Later calls spend that workspace's credits.

Inputs: workspace_id (required). Returns: object

talking_head_keyframeTalking head: keyframeAInspect

Step 1 of a talking head: the person on a set (close_up | waist_up | wide; landscape | portrait). Then text_to_speech, then talking_head_video.

Inputs: face_file (required), notes, format, framing, +3 more. Returns: object

talking_head_videoTalking head: animateAInspect

Animate a keyframe to a speech track (engine kling-ai-avatar-v2 or omnihuman-15; billed per second of speech, ≤ 60 s per segment).

Inputs: speech_file (required), keyframe_file (required), engine, motion_prompt, +1 more. Returns: object

text_to_speechText to speechCInspect

Generate speech (priced per 1,000 characters). voice from list_voices.

Inputs: text (required), model, voice, confirm_credits. Returns: object

try_onTry onCInspect

Fashion photo: the garments worn by the person in person_file, or by a described model (gender Female|Male|Non-binary, age 20s…60s+).

Inputs: garment_files (required), age, build, notes, +4 more. Returns: object

upload_fileUpload fileAInspect

Import a public http(s) image, video or audio URL into the workspace and get a file id to pass in a model's input slots. For a file with no public URL (a chat attachment, something on the user's device) use request_upload instead.

Inputs: url (required), filename. Returns: object

wait_for_generationWait for generationA
Read-only
Inspect

Block up to seconds (max 25) until the generation finishes, then return it. Call repeatedly for long video renders.

Inputs: generation_id (required), seconds. Returns: object

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    Try in Browser

    Glama MCP Gateway

    Add one secure layer between your agents and this server.

    Resources