microsoft/semantic-kernel · error · NotSupportedException

Unsupported audio format '{mimeType}'. Supported formats are

Error message

Unsupported audio format '{mimeType}'. Supported formats are 'audio/wav' and 'audio/mp3'.

What it means

Thrown by GetChatInputAudioFormat. For input audio the connector maps the mime-type to OpenAI's ChatInputAudioFormat, accepting only 'audio/wav' and 'audio/mp3'. Any other mime-type (ogg, flac, m4a, etc.) is not representable as an input-audio format and is rejected with an explicit list of allowed values.

Source

Thrown at dotnet/src/Connectors/Connectors.OpenAI/Core/ClientCore.ChatCompletion.cs:936

        {
            return ChatMessageContentPart.CreateFilePart(BinaryData.FromBytes(data), binaryContent.MimeType, Guid.NewGuid().ToString());
        }

        throw new ArgumentException($"{nameof(BinaryContent)} must have Data bytes.");
    }

    private static ChatInputAudioFormat GetChatInputAudioFormat(string? mimeType)
    {
        if (string.IsNullOrWhiteSpace(mimeType))
        {
            return ChatInputAudioFormat.Mp3;
        }

        return mimeType.ToUpperInvariant() switch
        {
            "AUDIO/WAV" => ChatInputAudioFormat.Wav,
            "AUDIO/MP3" => ChatInputAudioFormat.Mp3,
            _ => throw new NotSupportedException($"Unsupported audio format '{mimeType}'. Supported formats are 'audio/wav' and 'audio/mp3'.")
        };
    }

    private static ChatImageDetailLevel? GetChatImageDetailLevel(ImageContent imageContent)
    {
        const string DetailLevelProperty = "ChatImageDetailLevel";

        if (imageContent.Metadata is not null &&
            imageContent.Metadata.TryGetValue(DetailLevelProperty, out object? detailLevel) &&
            detailLevel is not null)
        {
            if (detailLevel is string detailLevelString && !string.IsNullOrWhiteSpace(detailLevelString))
            {
                return detailLevelString.ToUpperInvariant() switch
                {
                    "AUTO" => ChatImageDetailLevel.Auto,
                    "LOW" => ChatImageDetailLevel.Low,
                    "HIGH" => ChatImageDetailLevel.High,

View on GitHub (pinned to c028a0c7dc)

Solutions

  1. Transcode the source to WAV or MP3 and set MimeType to 'audio/wav' or 'audio/mp3'.
  2. If MimeType is missing, the connector defaults to Mp3 — supply bytes that are actually MP3 in that case.
  3. Validate the mime-type against the allowed set before constructing AudioContent.

Example fix

// before
var audio = new AudioContent(bytes, "audio/ogg");
// after
var audio = new AudioContent(wavBytes, "audio/wav");
Defensive patterns

Strategy: validation

Validate before calling

static readonly HashSet<string> ValidInputAudioMime = new(StringComparer.OrdinalIgnoreCase) { "audio/wav", "audio/mp3" };
static string NormalizeInputAudioMime(string? m) => string.IsNullOrWhiteSpace(m) ? "audio/mp3" : ValidInputAudioMime.Contains(m) ? m.ToLowerInvariant() : throw new ArgumentException($"Transcode to WAV/MP3; got {m}");

Type guard

static bool IsValidInputAudioMime(string? m) => string.IsNullOrWhiteSpace(m) || ValidInputAudioMime.Contains(m);

Try / catch

try { await client.GetChatCompletionAsync(history); }
catch (NotSupportedException ex) when (ex.Message.Contains("audio format")) { /* transcode to wav/mp3 and retry */ }

Prevention

When it happens

Trigger: Building an AudioContent for chat input with a MimeType other than audio/wav or audio/mp3 (e.g. 'audio/ogg', 'audio/m4a', or a null/whitespace handled separately that still resolves elsewhere).

Common situations: Recording in a container format the model's realtime/audio input does not accept; mismatched file extension vs actual codec; passing a generic 'application/octet-stream'.

Related errors

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/199b4ba1cab0c4ed. Report an issue: GitHub.