cinemap_find_assets
ChatGPTBrowse the user's Cinemap asset library, newest first: rendered takes, whitebox previews, reference and generated images, frame grabs, 3D models and audio. Filter with category or favorites_only. Each asset has an asset_id; pass it to cinemap_get_asset for a download link. When more assets exist the result carries next_cursor — pass it back as cursor for the next page. Read-only.
cinemap_get_asset
ChatGPTGet a short-lived download link for one asset in the user's Cinemap library. Pass an asset_id returned by cinemap_find_assets. The link expires (see expires_at); fetch a new one if it has. To use an image asset as a reference in TapNow production tools, pass the link to upload_image. Read-only.
cinemap_import_reference
ChatGPTCopy an image or video from TapNow into the user's Cinemap asset library, so it can be used as a reference while building scenes in Cinemap. Pass the image_id and project_id returned by a TapNow tool (for example upload_image or a production tool). Optionally pass world_id from cinemap_list_worlds to file it under that world. Importing the same file twice does not create a duplicate. Free.
cinemap_list_worlds
ChatGPTList the user's Cinemap worlds (3D scenes for previs and shot planning), most recently updated first. Each world has a world_id and, when available, a world_url that opens it in Cinemap. Pass keyword to filter by world name. Use this when the user asks about their Cinemap worlds or wants to continue one. Read-only.
create_audio
ChatGPTGenerate one audio clip — a voiceover, a dialogue scene, music, a song or a sound effect — with a model of the user's choice. Music and sound effects can also score a finished video. The output is saved to the user's TapNow canvas automatically — the caller does not need to know about canvases. Charges the user's account; longer audio costs more. Free TapNow accounts can make speech and sound effects and score a finished video; dialogue, music from a prompt and songs need a paid plan. Before calling, check whether the user has said what they want for the options marked 'ask the user' below. If they have not, ask once, in one short message that offers those choices; if they leave it to you, omit those parameters so the defaults apply. This is a job-ticket tool: it returns node_ids immediately while the audio is still being generated, so never tell the user it is ready at this point. Poll get_production_result with the returned node_ids until status is 'done'. The result's applied_params lists the settings actually used. Modes: - speech: one voice reading the prompt aloud, word for word (voiceover, narration). Models: elevenlabs-v3 (default). - dialogue: a scene rendered from a script: several speakers, with music and effects underneath (audio drama, podcast-style exchange). Models: seed-audio-1.0 (default). - music: instrumental or background music; with video_id, a score for that video. Models: sonilo (default), elevenlabs-v3. - song: a song with vocals, sung to the lyrics you give or ones the model writes. Models: elevenlabs-v3 (default). - sound_effect: a sound effect; with video_id, effects that follow that video. Models: sonilo (default), elevenlabs-v3. Models: - elevenlabs-v3: ElevenLabs: expressive voiceover in 20 voices, songs with your lyrics, instrumental music and short loopable sound effects. Modes: speech, music (3-300s), song (3-300s), sound_effect (1-30s). - seed-audio-1.0: Seed Audio 1.0: a whole scene from a script — several speakers with music and effects underneath, up to 120 seconds. Modes: dialogue. - sonilo: Sonilo: instrumental music up to 6 minutes and sound effects up to 3 minutes; the one that can score an existing video. Modes: music (5-360s), sound_effect (1-180s).
create_deck_visuals
ChatGPTTurn an outline into a set of style-consistent illustrations — one per section — for slides, documents or business proposals. All images share one visual language. Images never contain text (models render CJK text unreliably); space is left for the user to overlay copy themselves — say so before submitting. This is a job-ticket tool: it returns node_ids immediately while the images are still being generated, so never tell the user they are ready at this point. Poll get_production_result until status is 'done'. Charges the user's account per image.
create_hero_image
ChatGPTGenerate one high-quality image from a text description. The output is saved to the user's TapNow canvas automatically — the caller does not need to know about canvases. This is a job-ticket tool: it returns node_ids immediately while the image is still being generated, so never tell the user the image is ready at this point. Poll get_production_result with the returned node_ids until status is 'done'. Charges the user's account.
create_hero_video
ChatGPTGenerate one short video from a text description, optionally animating a still image. The output is saved to the user's TapNow canvas automatically — the caller does not need to know about canvases. Video generation costs significantly more than image generation, so tell the user this is a paid video render and get their agreement before calling. This is a job-ticket tool: it returns node_ids immediately while the video is still being rendered, so never tell the user the video is ready at this point. Video renders take minutes, not seconds — poll get_production_result with the returned node_ids until status is 'done'. Charges the user's account.
create_image
ChatGPTGenerate images with a model of the user's choice, from a text description and optionally up to 20 reference images to edit, restyle or combine. The output is saved to the user's TapNow canvas automatically — the caller does not need to know about canvases. This is a job-ticket tool: it returns node_ids immediately while the images are still being generated, so never tell the user they are ready at this point. Poll get_production_result with the returned node_ids until status is 'done'. Charges the user's account; the model, resolution, quality and count change the price, and free TapNow accounts can use seedream-5-lite only. Before calling, check whether the user has said what they want for the options marked 'ask the user' below. If they have not, ask once, in one short message that offers those choices; if they leave it to you, omit those parameters so the defaults apply. The result's applied_params lists the settings actually used. Models. Each accepts only the options listed for it; leave an option out to use the model's default (marked): - seedream-5-lite: General-purpose images and reference-based edits at 2K/3K. aspect_ratio 1:1 (default)/4:3/3:4/16:9/9:16/3:2/2:3/21:9; resolution 2K (default)/3K; count 1-4; up to 14 reference images. - seedream-5-pro: Seedream 5.0 Pro: higher-fidelity images and reference-based edits. aspect_ratio 1:1 (default)/4:3/3:4/16:9/9:16/3:2/2:3/21:9; resolution 1K/2K (default); count 1-4; up to 10 reference images. - nano-banana-2: Fast images and edits; the one with 512P output and up to 20 references. aspect_ratio Auto (default)/1:1/9:16/16:9/3:4/4:3/3:2/2:3/5:4/4:5/21:9; resolution 512P/1K (default)/2K/4K; count 1-4; up to 20 reference images. - nano-banana-2-lite: Lightweight, low-cost 1K images and edits. aspect_ratio Auto (default)/1:1/9:16/16:9/3:4/4:3/3:2/2:3/5:4/4:5/21:9; resolution 1K (default); count 1-4; up to 20 reference images. - nano-banana-2.1: Next-generation fast images and edits up to 4K, including 1:8 to 8:1 ratios. aspect_ratio Auto (default)/1:1/9:16/16:9/3:4/4:3/3:2/2:3/5:4/4:5/21:9/1:4/4:1/1:8/8:1; resolution 1K (default)/2K/4K; count 1-4; up to 14 reference images. - nano-banana-pro: Polished high-resolution images for refined visual design. aspect_ratio Auto (default)/1:1/9:16/16:9/3:4/4:3/3:2/2:3/5:4/4:5/21:9; resolution 1K/2K (default)/4K; count 1-4; up to 20 reference images. - gpt-image-2: Instruction-following images and edits for text, diagrams and layouts. aspect_ratio auto/1:1 (default)/4:3/3:4/5:4/4:5/3:2/2:3/16:9/9:16/21:9/9:21; resolution 1K (default)/2K/4K/auto; quality low (default)/medium/high; count 1-4; up to 16 reference images. - gpt-image-2.5-flare: Faster GPT Image 2.5 for readable text, layouts and reference-aware edits. aspect_ratio auto/1:1 (default)/4:3/3:4/5:4/4:5/3:2/2:3/16:9/9:16/21:9/9:21; resolution 1K (default)/2K/4K/auto; quality low (default)/medium/high/xhigh/max; count 1-4; up to 16 reference images. - gpt-image-2.5-sunburst: High-control GPT Image 2.5 for multi-step edits and instruction-heavy visuals. aspect_ratio auto/1:1 (default)/4:3/3:4/5:4/4:5/3:2/2:3/16:9/9:16/21:9/9:21; resolution 1K (default)/2K/4K/auto; quality low (default)/medium/high/xhigh/max; count 1-4; up to 16 reference images. - midjourney-v7: Midjourney v7 for aesthetic exploration; always 4 candidates. aspect_ratio 1:1/16:9 (default)/9:16/4:3/3:4/2:3/3:2/7:4/4:7; always 4 images; up to 4 reference images. - midjourney-v8.1: Faster Midjourney aesthetics; always 4 candidates. aspect_ratio 1:1/16:9 (default)/9:16/4:3/3:4/2:3/3:2/7:4/4:7; always 4 images; up to 4 reference images. - midjourney-v8.2: Midjourney v8.2 for aesthetic exploration, with optional HD; always 4 candidates. aspect_ratio 1:1/16:9 (default)/9:16/4:3/3:4/2:3/3:2/7:4/4:7; quality standard/hd; always 4 images; up to 4 reference images.
create_product_shots
ChatGPTTurn one product photo into a full set of e-commerce shots — white background hero, lifestyle scene, alternate angle and detail close-up — sized for the target platform. Requires a reference image: pass the image_id returned by upload_image. This is a job-ticket tool: it returns node_ids immediately while the images are still being generated, so never tell the user they are ready at this point. Poll get_production_result until status is 'done'. Charges the user's account per image.
create_video
ChatGPTGenerate one short video with a model of the user's choice, from a text description and optionally from still images (animate an image, go from a first to a last frame, or use reference images, up to the number each model allows). The output is saved to the user's TapNow canvas automatically — the caller does not need to know about canvases. Video generation costs significantly more than image generation, so tell the user this is a paid video render and get their agreement before calling; the model, resolution, quality, duration and sound change the price. Free TapNow accounts can use seedance-2.0-mini only, at 480p and 4-5 seconds. Before calling, check whether the user has said what they want for the options marked 'ask the user' below. If they have not, ask once, in one short message that offers those choices; if they leave it to you, omit those parameters so the defaults apply. This is a job-ticket tool: it returns node_ids immediately while the video is still being rendered, so never tell the user the video is ready at this point. Video renders take minutes, not seconds — poll get_production_result with the returned node_ids until status is 'done'. The result's applied_params lists the settings actually used. Models. Each accepts only the options and image modes listed for it; leave an option out to use the model's default (marked): - seedance-2.0-mini: Lightweight Seedance previews at 480p/720p. aspect_ratio 16:9 (default)/9:16/1:1/4:3/3:4/21:9; resolution 480p (default)/720p; duration 4-15s; generate_audio; with images: image_to_video/first_last_frame/reference. - seedance-2.0-fast: Fast Seedance previews at 480p/720p. aspect_ratio 16:9 (default)/9:16/1:1/4:3/3:4/21:9; resolution 480p/720p (default); duration 4-15s; generate_audio; with images: image_to_video/first_last_frame/reference. - seedance-2.0: Balanced default for general video, up to 4K. aspect_ratio 16:9 (default)/9:16/1:1/4:3/3:4/21:9; resolution 480p/720p (default)/1080p/4k; duration 4-15s; generate_audio; with images: image_to_video/first_last_frame/reference. - seedance-2.5: Long-form Seedance, 4-30 seconds, up to 30 reference images. aspect_ratio 16:9/9:16/1:1/4:3/3:4/21:9/adaptive (default); resolution 480p/720p (default)/1080p; duration 4-30s; generate_audio; with images: image_to_video/first_last_frame/reference. - seedance-2.5-draft: 480p Seedance 2.5 preview at a lower price. aspect_ratio 16:9/9:16/1:1/4:3/3:4/21:9/adaptive (default); resolution 480p (default); duration 4-30s; generate_audio; with images: image_to_video/first_last_frame/reference. - kling-video-o3: Kling: cinematic motion with std, pro and 4k quality modes. aspect_ratio 16:9 (default)/9:16/1:1; quality std/pro (default)/4k; duration 3-15s; generate_audio; with images: image_to_video/first_last_frame/reference. - wan-2.6: Wan 2.6 with sound and multi-shot support. aspect_ratio 16:9 (default)/9:16/1:1; resolution 720p/1080p; duration 5/10/15s (text_to_video/image_to_video), 2-10s (reference); generate_audio; with images: image_to_video/reference. - wan-3.0: Wan 3.0, 2-30 seconds, multi-image reference. aspect_ratio adaptive (default)/16:9/9:16/1:1/4:3/3:4; resolution 480p/720p/1080p; duration 2-30s; generate_audio; with images: image_to_video/reference. - happyhorse-1.0: HappyHorse for text, image and multi-image reference video. aspect_ratio 16:9 (default)/9:16/1:1/4:3/3:4; resolution 720p/1080p; duration 3-15s; with images: image_to_video/reference. - happyhorse-1.1: HappyHorse 1.1 for text, image and multi-image reference video. aspect_ratio 16:9 (default)/9:16/1:1/4:3/3:4; resolution 720p/1080p; duration 3-15s; with images: image_to_video/reference. - minimax-h3: MiniMax Hailuo-03 at 2K, with first/last frame and references. aspect_ratio 21:9/16:9 (default)/4:3/1:1/3:4/9:16; resolution 2k; duration 5-15s; with images: image_to_video/first_last_frame/reference. - minimax-h3-max: MiniMax H3 Max at 480P/768P from text or one image. aspect_ratio 21:9/16:9 (default)/4:3/1:1/3:4/9:16; resolution 480P/768P (default); duration 5-15s; with images: image_to_video. - flux-3: FLUX 3 vid…
find_library_assets
ChatGPTSearch or browse library assets and return compact paths with IDs.
find_library_folders
ChatGPTFind, resolve, or browse library folders and return compact paths with folder_id.
get_production_result
ChatGPTPoll the result of a TapNow production task by the node_ids returned when the task was submitted; pass back the project_id you received alongside them. Works for both image and video tasks. While status is 'generating', wait 15-30 seconds for images or 30-60 seconds for videos, then call again. When an item reaches status 'done' it carries a media_url that can be rendered or downloaded directly, and a media_type of 'image' or 'video'. Items with status 'failed' will not recover — stop polling and offer to resubmit. This tool never starts or charges a generation; when a result is ready it also records its media URL and File Hub identity on the canvas node so the image or video can be reused by other production tools. Items may also carry a task_id: that is a bookkeeping reference for the generation ledger, not a handle — always poll by node_id and never pass a task_id as a node_id.
list_visual_projects
ChatGPTList the user's TapNow canvas projects, most recently updated first, including team projects shared in the user's current workspace. Each project carries scope (private or team). Pass keyword to search by project name, or scope to list only private or only team projects. Use this when the user asks what they have been working on, or wants to continue an earlier project. Read-only, no side effects.
resize_for_platform
ChatGPTOpen a panel that reframes one existing canvas image into the exact aspect ratios different platforms need — WeChat article cover, slide, Xiaohongshu, Moments, Douyin, Taobao. The user picks which platforms they want and applies them in one go. Cropping is free: it never runs a model and never charges the user, so offer it freely whenever an image needs to fit somewhere specific. Pass only the image; the panel works out the framing itself and the user can adjust it.
resize_for_platform_apply
ChatGPTRender the platform crops the user selected in the resize panel and add one image per platform to the project, keeping the original untouched. Free: it never runs a model. The panel calls this itself — call it directly only when the user asked for specific formats and explicitly did not want the panel.
save_to_library
ChatGPTSave images that are already in TapNow into the user's asset library so they can be reused later and found with find_library_assets. Pass the image_id values returned by upload_image or the node_ids returned by production tools. Without folder_name the items go to a default channel folder. Read-only with respect to billing: this tool never charges the user.
show_library
ChatGPTOpen the interactive library picker panel so the user can browse folders, search assets, and pick one themselves. Use this when the user wants to choose something from their library but has not named an exact asset.
tapnow_status
ChatGPTCheck whether the TapNow MCP tool execution path is available. This tool is read-only, deterministic, and has no side effects.
upload_image
ChatGPTBring an existing image into TapNow from a public HTTPS URL so it can be used as input for later production tools. Use this when the user supplies a picture — from the conversation, a document, or anywhere else — and wants to generate from it. Returns an image_id that production tools accept as a reference image. Read-only with respect to billing: this tool never charges the user.
