{
  "title": "Audio file input",
  "description": "Upload recorded audio for use as model input.",
  "slug": "audio-file-input",
  "locale": "en",
  "seoTitle": "Audio file input compatibility — Can My Agent Use",
  "socialTitle": "Audio file input",
  "socialDescription": "Compare recorded-audio uploads, transcription support, and file limits.",
  "llmSummary": "Audio file input covers recorded uploads and is separate from realtime voice. Transcription, timestamps, speaker labels, and acoustic analysis are recorded as qualifiers only when documented.",
  "audience": "Engineers comparing recorded-audio and meeting workflows.",
  "contentKind": "feature",
  "status": "published",
  "tags": [
    "perception",
    "audio",
    "uploads",
    "transcription"
  ],
  "updated": "2026-08-28T00:00:00.000Z",
  "published": "2026-08-28T00:00:00.000Z",
  "category": "perception",
  "summary": "Upload recorded audio for use as model input.",
  "specLabel": "Common product term",
  "aliases": [
    "audio input",
    "audio upload",
    "recording attachment",
    "speech transcription"
  ],
  "capabilityKind": "atomic",
  "parent": "file-inputs",
  "related": [
    "video-input",
    "realtime-voice",
    "upload-limits"
  ],
  "relations": [],
  "highlight": false,
  "notes": [
    {
      "id": 1,
      "text": "Evidence checked 2026-08-28: xAI documents direct audio uploads in Grok chats and describes transcription and interpretation of audio and video inputs; detailed formats, timing, diarization, and duration limits are not stated on the reviewed FAQ page."
    },
    {
      "id": 2,
      "text": "Evidence checked 2026-08-28: Grok Bot lists audio among common supported inputs and documents a 25 MB per-audio-file limit, but the reviewed page does not specify transcript, speaker, timing, or acoustic fidelity."
    },
    {
      "id": 3,
      "text": "Evidence checked 2026-08-28: Gemini Apps accept audio uploads and document total audio-duration limits of 10 minutes without a Google AI plan or 3 hours with Google AI Pro or Ultra. The reviewed page does not define diarization, timestamp, transcript, or acoustic-understanding fidelity."
    },
    {
      "id": 4,
      "text": "Evidence checked 2026-08-28: Perplexity web file uploads accept MPEG, WAV, AIFF, OGG, FLAC, and MP3 audio, automatically transcribe spoken content, and can identify and label speakers. The general file-upload page documents a 40 MB per-file limit."
    },
    {
      "id": 5,
      "text": "Evidence checked 2026-08-28: Gemini CLI custom commands encode a supported audio path referenced with @{...} and inject it as multimodal input. The reviewed page does not enumerate audio formats or establish transcript, timestamp, diarization, or acoustic-analysis fidelity."
    }
  ],
  "issues": [],
  "resources": [
    {
      "title": "Methodology",
      "href": "/methodology",
      "kind": "note"
    },
    {
      "id": "xai-grok-files-faq",
      "title": "xAI — Grok files and data FAQ",
      "href": "https://docs.x.ai/grok/faq",
      "kind": "docs",
      "publisher": "xAI",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28",
      "locator": "Files & Data"
    },
    {
      "id": "xai-grok-bot-files",
      "title": "xAI — Grok Bot files and results",
      "href": "https://docs.x.ai/grok-bot/files-and-results",
      "kind": "docs",
      "publisher": "xAI",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28",
      "locator": "Attach files"
    },
    {
      "id": "google-gemini-upload-files",
      "title": "Google Gemini Apps Help — Upload and analyze files",
      "href": "https://support.google.com/gemini/answer/14903178?co=GENIE.Platform%3DDesktop&hl=en",
      "kind": "docs",
      "publisher": "Google",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28",
      "locator": "Upload and analyze files; upload limits"
    },
    {
      "id": "perplexity-file-uploads",
      "title": "Perplexity Help Center — File uploads",
      "href": "https://www.perplexity.ai/help-center/en/articles/10354807-file-uploads",
      "kind": "docs",
      "publisher": "Perplexity",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28",
      "locator": "Audio files; supported file types and size"
    },
    {
      "id": "gemini-cli-custom-commands",
      "title": "Gemini CLI — Custom commands",
      "href": "https://geminicli.com/docs/cli/custom-commands/",
      "kind": "docs",
      "publisher": "Google",
      "evidenceType": "documented",
      "reviewedAt": "2026-08-28",
      "locator": "Injecting file content with @{...}; multimodal support"
    }
  ],
  "support": [
    {
      "harness": "grok-web",
      "versions": [
        {
          "track": "current",
          "status": "yes",
          "noteIds": [
            1
          ],
          "target": {
            "kind": "hosted-observation",
            "revision": "2026-08-28 Grok Web documentation observation",
            "observedAt": "2026-08-28"
          },
          "environmentProfile": "hosted-default",
          "qualifiers": [
            {
              "type": "runtime",
              "value": "transcription and interpretation are documented, but accepted formats, diarization, timestamp, and native-audio fidelity are not specified on the reviewed page"
            }
          ],
          "evidence": [
            {
              "resourceId": "xai-grok-files-faq",
              "type": "documented",
              "observedAt": "2026-08-28"
            }
          ]
        }
      ]
    },
    {
      "harness": "grok-bot-desktop",
      "versions": [
        {
          "track": "current",
          "status": "yes",
          "noteIds": [
            2
          ],
          "target": {
            "kind": "hosted-observation",
            "revision": "2026-08-28 Grok Bot desktop documentation observation",
            "observedAt": "2026-08-28"
          },
          "environmentProfile": "hosted-default",
          "qualifiers": [
            {
              "type": "runtime",
              "value": "audio is a supported attachment up to 25 MB, but transcript, speaker, timing, and acoustic semantics are not documented on the reviewed page"
            }
          ],
          "evidence": [
            {
              "resourceId": "xai-grok-bot-files",
              "type": "documented",
              "observedAt": "2026-08-28"
            }
          ]
        }
      ]
    },
    {
      "harness": "gemini-web",
      "versions": [
        {
          "track": "current",
          "status": "yes",
          "noteIds": [
            3
          ],
          "target": {
            "kind": "hosted-observation",
            "revision": "2026-08-28 Gemini Apps documentation observation",
            "observedAt": "2026-08-28"
          },
          "environmentProfile": "hosted-default",
          "qualifiers": [
            {
              "type": "runtime",
              "value": "direct audio upload and analysis are documented"
            },
            {
              "type": "plan",
              "value": "total audio is limited to 10 minutes without a Google AI plan and 3 hours with Google AI Pro or Ultra"
            },
            {
              "type": "runtime",
              "value": "accepted audio formats, diarization, timestamps, transcript fidelity, and native acoustic semantics are not established by the reviewed page"
            }
          ],
          "evidence": [
            {
              "resourceId": "google-gemini-upload-files",
              "type": "documented",
              "observedAt": "2026-08-28"
            }
          ]
        }
      ]
    },
    {
      "harness": "perplexity-web",
      "versions": [
        {
          "track": "current",
          "status": "yes",
          "noteIds": [
            4
          ],
          "target": {
            "kind": "hosted-observation",
            "revision": "2026-08-28 Perplexity web documentation observation",
            "observedAt": "2026-08-28"
          },
          "environmentProfile": "hosted-default",
          "qualifiers": [
            {
              "type": "runtime",
              "value": "direct audio uploads are automatically transcribed; the documented formats are MPEG, WAV, AIFF, OGG, FLAC, and MP3, with a 40 MB general per-file limit"
            },
            {
              "type": "runtime",
              "value": "speaker identification and labels are documented, but timestamps and non-speech acoustic or prosody fidelity are not established"
            }
          ],
          "evidence": [
            {
              "resourceId": "perplexity-file-uploads",
              "type": "documented",
              "observedAt": "2026-08-28"
            }
          ]
        }
      ]
    },
    {
      "harness": "gemini-cli",
      "versions": [
        {
          "track": "current",
          "status": "yes",
          "noteIds": [
            5
          ],
          "target": {
            "kind": "dated-documentation",
            "revision": "current Gemini CLI custom-command documentation",
            "observedAt": "2026-08-28"
          },
          "environmentProfile": "local-default",
          "qualifiers": [
            {
              "type": "runtime",
              "value": "supported audio files referenced with @{...} inside a custom command are encoded and injected as multimodal input"
            },
            {
              "type": "runtime",
              "value": "accepted audio formats, ordinary-prompt attachment methods, transcripts, timestamps, diarization, and acoustic semantics are not established by the reviewed page"
            }
          ],
          "evidence": [
            {
              "resourceId": "gemini-cli-custom-commands",
              "type": "documented",
              "observedAt": "2026-08-28"
            }
          ]
        }
      ]
    }
  ],
  "html": "/features/audio-file-input",
  "markdown": "/features/audio-file-input.md",
  "json": "/api/v1/features/audio-file-input.json",
  "body": "This row covers recorded audio supplied as a file, not a live duplex voice session. Evidence should distinguish speech transcription from speaker diarization, timestamps, language identification, non-speech sound analysis, emotion or prosody claims, and native audio-model input.\n\nRecord formats, channels, maximum bytes and duration, language limits, transcript editability, retention, model and plan requirements, and whether tools or sub-agents receive the original recording or only a derivative transcript."
}