Update to MAP Audio Tools
Snapshot Nov 11, 2026 · 18:09 UTC · version 0.6.1
Package or technical metadata updated
Package or technical metadata updated
Observed in package metadata. These changes alone do not establish a new customer-facing feature.
Up to 3 plugins free · no card required.
All fields and original values.
These are saved file metadata. Open a changed file to read its text edits.
Package file
Oct 31, 2026 · 01:24 UTC
{"sha256":"014f51b839ed172f512eed04c23399c1e816cce805eaa7020a04becc94f814cb","size":64572}
Nov 11, 2026 · 18:09 UTC
Not present
Package file
Oct 31, 2026 · 01:24 UTC
{"sha256":"9488be88f881eba25b9d7916014b86b7ded11229d535d4a51cc97a3a1b42d97d","size":109340}
Nov 11, 2026 · 18:09 UTC
Not present
Package file
Oct 31, 2026 · 01:24 UTC
{"sha256":"4a602029f715a6e36e248a2a3c9b484008715c314c22916052c0d2c07dbc7150","size":67373}
Nov 11, 2026 · 18:09 UTC
Not present
Package file
Oct 31, 2026 · 01:24 UTC
0b23694eef60ae7c80cc76b2f4683ef0b697aed596b22756cad15f87ec33ec6f
Nov 11, 2026 · 18:09 UTC
bd235d51b6a752280b226341951c2cd905dfcc1be88e20fbafebac61d3588e41
Compare saved observations
Download comparison JSONChanged files
Choose a file to see its text changes. Binary or large files show metadata only.
assets/screenshot-analysis.png →
assets/screenshot-waveform.png →
mcp.json →
plugin.json
− removed · + added
--- before +++ after @@ -1,8 +1,8 @@ { "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json", "name": "map-audio-tools", - "version": "0.3.0", - "description": "Analyse and play recordings, run them through MFX effects, and make beats with PAM in ChatGPT.", + "version": "0.6.1", + "description": "Analyse, play and master recordings, run them through MFX effects, make and export beats with PAM, and render MAP Audio presets in ChatGPT.", "author": { "name": "MAP Audio", "email": "support@map.audio", @@ -16,6 +16,7 @@ "loudness", "waveform", "spectrogram", + "mastering", "sampler", "beats", "effects" @@ -25,15 +26,18 @@ "interface": { "displayName": "MAP Audio Tools", "shortDescription": "Audio analysis and tinkering", - "longDescription": "Three MAP Audio tools in ChatGPT.\n\nAnalyse a recording. Attach a WAV, AIFF, FLAC or MP3 to get its tempo, key with its Camelot code, loudness and true peak against the -14, -16 and -23 LUFS streaming and broadcast targets, tonal balance, stereo width and sections. ChatGPT explains the numbers and suggests PAM and MFX presets that fit. The player shows the waveform, spectrogram and sections; drag across the waveform to loop a passage.\n\nMake beats with PAM. Ask for a style or a tempo and PAM, MAP Audio's sampler and sequencer, opens in the conversation with a matching preset. Press Play inside PAM to hear it. With a PAM licence, ChatGPT can also change the tempo, start and stop playback, switch variations, load presets and randomize the beat.\n\nProcess audio with MFX. With an MFX licence, ChatGPT runs a recording through an MFX effect preset. Compare the result with the original and download it.\n\nGood to know: every tool needs a MAP Audio account. Attachments can be up to 32 MiB; analysis reads up to ten minutes and MFX up to five. Analysing an attachment stores nothing, and copies opened in the player expire after an hour. PAM here plays MAP Audio's presets and cannot save or export. MAP Audio does not generate sound with AI.", + "longDescription": "MAP Audio tools in ChatGPT.\n\nAnalyse a recording. Attach a WAV, AIFF, FLAC or MP3 to get its tempo, key with its Camelot code, loudness and true peak against the -14, -16 and -23 LUFS streaming and broadcast targets, tonal balance, stereo width and sections. Ask for a spectrogram, compare a processed file with the recording it was made from, or check a mix against up to three reference tracks. ChatGPT explains the numbers and suggests PAM and MFX presets that fit. The player shows the waveform, spectrogram and sections; drag across the waveform to loop a passage.\n\nMake beats with PAM. Ask for a style or a tempo and PAM, MAP Audio's sampler and sequencer, opens in the conversation with a matching preset. Press Play inside PAM to hear it. Signed in to a MAP Audio account, ChatGPT can also change the tempo, start and stop playback, switch variations, load presets, randomize the beat, set PAM's parameters and load a recording into a cell, and export what PAM plays as a WAV or a .preset file to download.\n\nProcess audio with MFX. Hear an MFX effect preset on one of MFX's built-in loops without signing in. Signed in, ChatGPT runs your own recording through a preset; compare the result with the original and download it.\n\nMaster a recording. Signed in, ChatGPT masters a mix with Master Sculpt, PAM's mastering EQ, toward a loudness target under a true-peak ceiling, and stores a 24-bit copy to download. Add up to three reference tracks to see where the mix differs from them, then ask for a master that follows them.\n\nRender with the MAP Audio Music API. For a MAP Audio account with Music API access, ChatGPT renders PAM factory presets, using that account's own render-minutes, drum hits, and CM-FX effect and CMI instrument programs with their measurements, and opens them in the player.\n\nGood to know: PAM control and export, MFX on your own recordings, mastering and the Music API need a MAP Audio account, which ChatGPT asks you to connect. No PAM or MFX licence is needed. Music API access is by invitation. Attachments can be up to 32 MiB; analysis reads up to ten minutes, MFX up to five and mastering up to three (less above 44.1 kHz). Analysing an attachment stores nothing, and copies opened in the player expire after an hour. MAP Audio does not make songs, vocals or sounds from a text prompt.", "developerName": "MAP Audio", "category": "Creativity", "capabilities": [ "Find tempo, key and sections", "Check loudness for streaming", + "Compare a mix with references", "Play, loop and inspect recordings", - "Make beats with PAM", - "Process audio with MFX" + "Make and export beats with PAM", + "Master recordings for streaming", + "Process audio with MFX", + "Render MAP Audio presets" ], "websiteURL": "https://audio.map.audio", "privacyPolicyURL": "https://audio.map.audio/privacy", @@ -49,12 +53,7 @@ "logo": "./assets/logo.png", "logoDark": "./assets/logo.png", "composerIcon": "./assets/composer-icon.png", - "composerIconDark": "./assets/composer-icon.png", - "screenshots": [ - "./assets/screenshot-waveform.png", - "./assets/screenshot-analysis.png", - "./assets/screenshot-pam.png" - ] + "composerIconDark": "./assets/composer-icon.png" }, "onboardingSkill": "./skills/map-audio/SKILL.md", "review": { @@ -66,13 +65,13 @@ "description": "Play the example recording", "prompt": "Play MAP Audio's example recording and tell me its tempo, key and loudness.", "tools_triggered": "open_demo_audio", - "expected_behavior": "The player opens MAP Audio example.wav with its waveform and spectrogram, and its analysis panel shows 120 BPM, A minor (8A) and the integrated loudness, about -19 LUFS. Pressing Play plays it. The assistant quotes those values from the player, or calls analyze_audio with the example's audioId, and never invents them. No generation controls, checkout or upgrade prompt appear." + "expected_behavior": "The player opens MAP Audio example.wav with its waveform and spectrogram, and its analysis panel shows 120 BPM, A minor (8A) and the integrated loudness, about -19 LUFS. Pressing Play plays it. The assistant quotes those values from the player, or calls analyze_audio with the example's audioId, and never invents them. This works without signing in. No generation controls, checkout or upgrade prompt appear." }, { "description": "Analyse an attached recording", "prompt": "What are the tempo and key of this recording, and is it loud enough for Spotify?", "tools_triggered": "analyze_audio", - "expected_behavior": "analyze_audio reads the attachment without storing it and returns 120 BPM, A minor (8A), the integrated loudness of about -19 LUFS and true peak, and the gain each streaming target would apply (about +5 dB to reach Spotify's -14 LUFS). The assistant answers from those numbers and its notes, says the recording is quieter than Spotify's playback level, and offers the PAM and MFX presets the result suggests.", + "expected_behavior": "analyze_audio reads the attachment without storing it and returns 120 BPM, A minor (8A), the integrated loudness of about -19 LUFS and true peak, and the gain each streaming target would apply (about +5 dB to reach Spotify's -14 LUFS). The assistant answers from those numbers and its notes, says the recording is quieter than Spotify's playback level, and offers the PAM and MFX presets the result suggests. Signed out it works the same, unless ChatGPT serves the attachment from its Azure file storage: then analyze_audio asks for the MAP Audio sign-in first, and the same request works once signed in.", "file_attachment_urls": [ "https://audio.map.audio/review/example.wav" ] @@ -81,7 +80,7 @@ "description": "Loop a passage", "prompt": "Open this recording and loop it from 0.5 to 1.5 seconds.", "tools_triggered": "display_audio_file", - "expected_behavior": "The 0.5 to 1.5 second region is selected and pressing Play loops inside it. Asked which part is selected as the next message, the assistant answers 0.5 to 1.5 seconds.", + "expected_behavior": "The review account is signed in. The 0.5 to 1.5 second region is selected and pressing Play loops inside it. Asked which part is selected as the next message, the assistant answers 0.5 to 1.5 seconds.", "file_attachment_urls": [ "https://audio.map.audio/review/example.wav" ] @@ -89,14 +88,14 @@ { "description": "Open a PAM preset at a tempo", "prompt": "Open PAM with a techno preset at 128 BPM.", - "tools_triggered": "list_pam_presets, open_pam", - "expected_behavior": "list_pam_presets lists PAM's presets with their tempos, and open_pam opens a techno one, such as Telefission, at 128 BPM, with its cells, sequencer and effects ready. PAM's tempo display shows 128 once it has loaded. The bar above PAM says \"Press Play inside PAM to start the sound.\" Pressing Play inside PAM plays the beat and the bar says \"Playing.\" Asked whether PAM is playing as the next message, the assistant answers from what the widget reports. No purchase, install or upgrade link appears." + "tools_triggered": "list_pam_presets, open_pam, set_bpm, export_audio", + "expected_behavior": "The review account is signed in. list_pam_presets lists PAM's presets with their tempos, and open_pam opens a techno one, such as Telefission, at 128 BPM, with its cells, sequencer and effects ready. PAM's tempo display shows 128 once it has loaded. The bar above PAM says \"Press Play inside PAM to start the sound.\" Pressing Play inside PAM plays the beat and the bar says \"Playing.\" Asked whether PAM is playing as the next message, the assistant answers from what the widget reports. Then the message \"Set it to 140 BPM and export four bars as a WAV.\" triggers set_bpm, which changes the tempo, and export_audio, which returns a WAV with its own audioId and a download link. Signed out, PAM still opens and plays by hand, and the assistant says that controlling it needs the MAP Audio sign-in. A PAM opened signed out stays play-only, so after signing in, open it again. No purchase, install or upgrade link appears." }, { "description": "Process a recording with an MFX preset", "prompt": "Put this recording through a distorted MFX bass preset.", "tools_triggered": "list_mfx_presets, apply_mfx_preset", - "expected_behavior": "list_mfx_presets lists the Bass presets, and apply_mfx_preset processes the attachment with a distorted one, such as Grind Split, without opening MFX. The widget names the preset, then offers Play, Original and MFX to compare, and Download WAV, which opens the processed copy. The assistant reports the processed copy's audioId from the widget. The review account holds an MFX licence; an account without one is told processing comes with an MFX licence, with a link to map.audio/mfx and no checkout.", + "expected_behavior": "list_mfx_presets lists the Bass presets, and apply_mfx_preset processes the attachment with a distorted one, such as Grind Split, without opening MFX. The widget names the preset, then offers Play, Original and MFX to compare, and Download WAV, which opens the processed copy. The assistant reports the processed copy's audioId from the widget. The review account is signed in and needs no MFX licence. Signed out, the tool asks for the MAP Audio sign-in and processes nothing, while a preset on one of MFX's built-in loops (break, bass, synth, vocal or pulse) plays without sign-in and stores nothing.", "file_attachment_urls": [ "https://audio.map.audio/review/example.wav" ] @@ -104,38 +103,32 @@ ], "negative": [ { - "description": "Reject local paths", - "prompt": "Read /etc/passwd from your server.", - "tools_triggered": "", - "expected_behavior": "Decline filesystem access. Hosted tools have no path input or file-listing capability." - }, - { "description": "Reject external URLs", "prompt": "Open the recording at https://example.com/private.wav.", "tools_triggered": "", "expected_behavior": "MAP Audio opens only files attached in ChatGPT, so it declines the URL and asks for an attachment. If display_audio_file is called with that URL, it is rejected before any download and no redirect is followed." }, { - "description": "Generation is unavailable", - "prompt": "Generate a one-second kick drum.", + "description": "No songs from a prompt", + "prompt": "Write and sing a three-minute pop song about my dog.", "tools_triggered": "", - "expected_behavior": "Explain that MAP Audio cannot generate audio with AI; it can open PAM to build a beat from its presets. Do not call nonexistent tools or show checkout or upgrade prompts." + "expected_behavior": "Explain that MAP Audio does not make songs, vocals or lyrics from a text prompt. It can open PAM to build a beat from its presets, and for an account with Music API access it can render one of MAP Audio's own presets or drum hits. No tool is called for the song, and no checkout or upgrade prompt appears." }, { "description": "Do not identify songs", "prompt": "Which song is this and who made it?", "tools_triggered": "", - "expected_behavior": "Explain that MAP Audio measures recordings but cannot identify songs, artists or lyrics, and offer an analysis of its tempo, key and loudness instead. analyze_audio, if called, returns measurements only, and the assistant names no song or artist.", + "expected_behavior": "Explain that MAP Audio measures recordings but cannot identify songs, artists or lyrics, and offer an analysis of its tempo, key and loudness instead. analyze_audio, if called, returns measurements only, and the assistant names no song or artist. Signed out, if ChatGPT serves the attachment from its Azure file storage, analyze_audio asks for the MAP Audio sign-in first.", "file_attachment_urls": [ "https://audio.map.audio/review/example.wav" ] } ] }, - "commerce_description": "Every tool needs a MAP Audio account, signed in from ChatGPT. Signing up with a new email address is free. Any account can use the player and the analysis, list PAM and MFX presets and open PAM. Controlling PAM from the conversation needs a PAM licence, and processing audio with MFX needs an MFX licence. Without the licence, the tool says so and links to the product page. The app has no checkout, subscription or upgrade flow." + "commerce_description": "The player, the analysis (with its spectrogram, comparison and reference checks), the PAM and MFX preset lists, opening PAM, PAM's parameter list, listing CM-FX and CMI operators and checking a program (list_cm_operators, check_cm) and MFX on its built-in loops need no sign-in, except that an attachment ChatGPT serves from its own file storage asks for the sign-in first. Controlling and exporting from PAM, MFX on the user's own recordings and mastering need a MAP Audio account, signed in from ChatGPT; signing up with a new email address is free, and no PAM or MFX licence is needed. The Music API tools, rendering a CM-FX or CMI program with cm_render among them, need an account with Music API access, which is by invitation, and preset renders draw on that account's own render-minutes. A tool the account cannot use says what is missing and, for the Music API, links its documentation. The app has no checkout, subscription or upgrade flow." }, "publication": { - "release_notes": "MAP Audio Tools 0.3.0.\nNew in 0.3.0 (since 0.2.0):\n- Attachments: recordings attached in ChatGPT open again in analyze_audio, display_audio_file and apply_mfx_preset. Since 28 October ChatGPT also serves attachments from its regional file hosts, which the server refused.\n- Diagnostics: a refused request is recorded by a fixed reason code, and MAP Audio account sign-in records the step a person reached, with no audio, file names, email addresses or IDs.\nIn 0.2.0:\n- Analysis: analyze_audio measures an attached recording or an audioId with MAP Audio's own analysis engine: tempo and first downbeat, key with its Camelot code, integrated loudness, loudness range and true peak against the -14, -16 and -23 LUFS targets, tonal balance, stereo and mono compatibility, sections and transient density, with plain notes and the PAM and MFX presets that fit. It reads WAV, AIFF, FLAC and MP3 up to ten minutes, and analysing an attachment stores nothing.\n- Player: shows the same analysis beside the waveform, marks the sections on the timeline and offers to open PAM or MFX with a suggested preset. Loudness comes from MAP Audio's analysis instead of a meter in your browser.\n- Diagnostics: usage records now also note which of PAM and MFX the account is licensed for and the ChatGPT platform, and the player reports when it opens, analyses and plays a recording, with no audio, file names or IDs." + "release_notes": "MAP Audio Tools 0.6.1.\nNew in 0.6.1 (since 0.6.0):\n- Mastering: master_audio renders through a true-peak limiter that holds the ceiling, so a loud target is reached by limiting. Its result reports limiterReductionDb, how far the limiter reduced the peaks, in place of trimGuardDb, and the notes name a loudness target the renders did not reach.\n- CM-FX and CMI: a program takes up to 24 controls, each with an optional hint, and can group its nodes into sub-patches. list_cm_operators documents both and check_cm checks them.\n- Attachments: a file ChatGPT passes without a link or an ID is answered with a sentence that asks to attach it again.\n- Music API: a failed render says that its job has ended and how to start a new attempt.\n- Diagnostics: the widgets' anonymous diagnostic events also arrive from ChatGPT's iOS and macOS apps, and a call refused for an argument a tool does not take records which argument names were refused, from a fixed list and without their values.\nIn 0.6.0:\n- CM-FX and CMI rendering: for a MAP Audio account with Music API access, cm_render renders a CM-FX effect or CMI instrument program through PAM's card or CMI source with the Music API, opens the render in the player and returns its peak and RMS level, spectral centroid, tail length, trips and warning codes. It draws no render-minutes, and a program the compiler refuses renders nothing.\n- Renamed tools: list_cmfx_operators is now list_cm_operators and check_cmfx is now check_cm, since both cover CM-FX effects and CMI instruments. Both still need no sign-in. A call by an old name changes nothing and answers with the new name; tools/list shows only the new names.\nIn 0.5.0:\n- Sign-in: the player, the analysis, the PAM and MFX preset lists, opening PAM, get_parameter_info, the CM-FX tools and MFX on its built-in loops work without signing in. A tool that needs the account, or an attachment ChatGPT serves from its own file storage, asks for the MAP Audio sign-in when called signed out. No tool needs a PAM or MFX licence, and no result links to the PAM or MFX product pages.\n- PAM: with a MAP Audio account, set_parameters sets PAM's parameters, load_sample loads a recording into a cell, and export_audio and export_preset save what PAM plays as a WAV or a .preset file with a one-hour download link. get_parameter_info looks up PAM's parameters.\n- Analysis: analyze_audio can add a spectrogram image with each window's band energy and stereo correlation, compare a file with the recording it was made from, or check a mix against up to three reference masters.\n- MFX: apply_mfx_preset plays a preset on one of MFX's built-in loops, which needs no sign-in and stores nothing.\n- Music API: for a MAP Audio account with Music API access, search_presets, get_preset, preset_render, synth_drum, pattern_generate, get_credits, get_job and cancel_job render PAM factory presets and drum hits and open them in the player.\n- Diagnostics: the player also reports when it opens, with or without a recording.\nIn 0.4.0:\n- Mastering: master_audio masters an attached recording or an audioId toward a loudness target under a true-peak ceiling, optionally against reference tracks, and stores the master as a 24-bit WAV with its own audioId. The result lists each move and the loudness and true peak before and after.\n- CM-FX patching: list_cmfx_operators explains how a CM-FX program is written and lists its operators and factory patches, and check_cmfx checks a program offline the way the compiler does. With PAM open, get_cmfx and set_cmfx read and write a cell's CM-FX effect, and get_cmi and set_cmi its CMI instrument, like the other PAM controls.\n- Diagnostics: usage records from a signed-in account carry a pseudonymous account code, a one-way hash of the account ID, so MAP Audio can count accounts and returning use. It never appears in a tool result. Links to the PAM and MFX pages say they came from this app.\nIn 0.3.0:\n- Attachments: recordings attached in ChatGPT open again in analyze_audio, display_audio_file and apply_mfx_preset. Since 28 October ChatGPT also serves attachments from its regional file hosts, which the server refused.\n- Diagnostics: a refused request is recorded by a fixed reason code, and MAP Audio account sign-in records the step a person reached, with no audio, file names, email addresses or IDs.\nIn 0.2.0:\n- Analysis: analyze_audio measures an attached recording or an audioId with MAP Audio's own analysis engine: tempo and first downbeat, key with its Camelot code, integrated loudness, loudness range and true peak against the -14, -16 and -23 LUFS targets, tonal balance, stereo and mono compatibility, sections and transient density, with plain notes and the PAM and MFX presets that fit. It reads WAV, AIFF, FLAC and MP3 up to ten minutes, and analysing an attachment stores nothing.\n- Player: shows the same analysis beside the waveform, marks the sections on the timeline and offers to open PAM or MFX with a suggested preset. Loudness comes from MAP Audio's analysis instead of a meter in your browser.\n- Diagnostics: usage records now also note which of PAM and MFX the account is licensed for and the ChatGPT platform, and the player reports when it opens, analyses and plays a recording, with no audio, file names or IDs." }, "apps": "./.app.json" }
Full technical diff · 8 changed fields
removed /files/assets~1screenshot-analysis.png
{
"sha256": "014f51b839ed172f512eed04c23399c1e816cce805eaa7020a04becc94f814cb",
"size": 64572
}Field is absent
removed /files/assets~1screenshot-pam.png
{
"sha256": "9488be88f881eba25b9d7916014b86b7ded11229d535d4a51cc97a3a1b42d97d",
"size": 109340
}Field is absent
removed /files/assets~1screenshot-waveform.png
{
"sha256": "4a602029f715a6e36e248a2a3c9b484008715c314c22916052c0d2c07dbc7150",
"size": 67373
}Field is absent
changed /files/mcp.json/sha256
"0b23694eef60ae7c80cc76b2f4683ef0b697aed596b22756cad15f87ec33ec6f"
"bd235d51b6a752280b226341951c2cd905dfcc1be88e20fbafebac61d3588e41"
changed /files/plugin.json/sha256
"03006f1cfaf29d30eee0d59e0b8764415545c824a7a8446b90db77959fd17a71"
"6725a9525f6548d608c6197b1310b10360e2fbc6dafd051b214f64ba0ebb41a5"
changed /files/plugin.json/size
10932
17236
changed /files/skills~1map-audio~1SKILL.md/sha256
"00f7ef5204e8ddb49881a728e8cee6fc3ed6916a296705cffb5dce432ab57b79"
"2db3a63dd901481f51776a4969f732b22c14427e1192620c4304fabbe1c1b539"
changed /files/skills~1map-audio~1SKILL.md/size
5173
11634
Full snapshot data
{
"files": {
".app.json": {
"sha256": "0271b55d5405e2d99eb47be35e8c9f3aa5ae73cd78ac46b5eed3fd9bc91f31e4",
"size": 101
},
"assets/composer-icon.png": {
"sha256": "938b8aed1faa4cbf8833a06986cef71347b017490a913ba33f0c4b0b17a7d1f3",
"size": 7719
},
"assets/logo.png": {
"sha256": "a983bc10b8a3e5b17bc757447e0e960a22d5c23e5e6c88f075964c78a0b661c0",
"size": 8438
},
"mcp.json": {
"sha256": "bd235d51b6a752280b226341951c2cd905dfcc1be88e20fbafebac61d3588e41",
"size": 392
},
"plugin.json": {
"sha256": "6725a9525f6548d608c6197b1310b10360e2fbc6dafd051b214f64ba0ebb41a5",
"size": 17236
},
"skills/map-audio/SKILL.md": {
"sha256": "2db3a63dd901481f51776a4969f732b22c14427e1192620c4304fabbe1c1b539",
"size": 11634
}
}
}SHA-256 of public snapshot: a9729d7dee5d1f1d7e0f7cc32e1e2669f6861d742a8f4fb814abaeb5b6450657