vercel/ai · error · UnsupportedFunctionalityError

non-streaming transcription with ${this.modelId}

Error message

non-streaming transcription with ${this.modelId}

What it means

The ElevenLabs transcription model throws this when doGenerate (non-streaming transcription) is called with a model ID that is a realtime-only transcription model (per isRealtimeTranscriptionModelId). Realtime models like scribe_v1_realtime only support WebSocket streaming; they cannot be used for one-shot HTTP transcription.

Source

Thrown at packages/elevenlabs/src/elevenlabs-transcription-model.ts:185

            key as keyof ElevenLabsTranscriptionAPITypes
          ];
        if (value !== undefined) {
          formData.append(key, String(value));
        }
      }
    }

    return {
      formData,
      warnings,
    };
  }

  async doGenerate(
    options: Parameters<TranscriptionModelV4['doGenerate']>[0],
  ): Promise<Awaited<ReturnType<TranscriptionModelV4['doGenerate']>>> {
    if (isRealtimeTranscriptionModelId(this.modelId)) {
      throw new UnsupportedFunctionalityError({
        functionality: `non-streaming transcription with ${this.modelId}`,
      });
    }

    const currentDate = this.config._internal?.currentDate?.() ?? new Date();
    const { formData, warnings } = await this.getArgs(options);

    const {
      value: response,
      responseHeaders,
      rawValue: rawResponse,
    } = await postFormDataToApi({
      url: this.config.url({
        path: '/v1/speech-to-text',
        modelId: this.modelId,
      }),
      headers: combineHeaders(this.config.headers?.(), options.headers),
      formData,

View on GitHub (pinned to 69428b1f8b)

Solutions

Use a batch/non-realtime model id such as 'scribe_v1' for non-streaming transcription.
If you need realtime results, use doStream with the realtime model instead of doGenerate.
Check the model id with isRealtimeTranscriptionModelId-style logic before choosing the call path.

Example fix

// before
const model = elevenlabs.transcriptionModel('scribe_v1_realtime');
await model.doGenerate(options);
// after
const model = elevenlabs.transcriptionModel('scribe_v1');
await model.doGenerate(options);

Defensive patterns

Strategy: validation

Validate before calling

const REALTIME_IDS = ['scribe_v1_realtime'];
function assertBatchModel(modelId: string) {
  if (REALTIME_IDS.some(id => modelId.includes(id))) {
    throw new Error(`${modelId} only supports streaming transcription`);
  }
}
assertBatchModel('scribe_v1'); // before calling doGenerate

Type guard

function isRealtimeModelId(modelId: string): boolean {
  return modelId.includes('realtime');
}

Prevention

Keep separate factory helpers for batch vs realtime model ids.
Centralize model id selection in one config module with capability flags.
Read the ElevenLabs model docs before switching model ids.

When it happens

Trigger: Calling generateText/transcribe-style doGenerate with providerOptions selecting a realtime model id such as 'scribe_v1_realtime' through elevenlabs.transcriptionModel('scribe_v1_realtime').

Common situations: Developers copy a realtime model id from ElevenLabs docs into a non-streaming transcription call, or dynamically switch model ids and accidentally route a realtime model into the batch transcription path.

Related errors

AI-assisted analysis of vercel/ai@69428b1f8b (2026-08-30). Data as JSON: /api/errors/0430458efc38e480. Report an issue: GitHub.