ASSEMBLYAI ALTERNATIVE
Like AssemblyAI, plus every other modality.
Transcribe with Whisper through the OpenAI audio endpoint, then summarize or tag the transcript with any chat model, all on one key.
Hand this to your agent
Add a Runcrate model to this project: run curl -fsSL"https://runcrate.ai/agents.md?intent=build"and follow it.
Your agent asks for an API key. Create a key
Comparison
Runcrate and AssemblyAI, side by side.
| Feature | Runcrate | AssemblyAI |
|---|---|---|
| Transcription price | Whisper Large v3 Turbo: $0.00021 per minute. Voxtral Small: $0.00315 per minute | Universal-2: $0.15/hr ($0.0025/min). Universal-3.5 Pro: $0.21/hr ($0.0035/min) |
| Live audio | Chunked HTTP requests, billed per second of audio | Universal-Streaming $0.15/hr; Universal-3.6 Pro Realtime $0.45/hr |
| Speaker diarization | Not built in | $0.02/hr add-on on async audio |
| Languages | Whisper's 99, plus Cantonese | 99 on Universal-2, 18 on Universal-3.5 Pro |
| Summaries, topics, sentiment | Any chat model on the transcript, same key | Its Speech Understanding and LLM Gateway products |
| API format | OpenAI-compatible /v1/audio/transcriptions | AssemblyAI's own API |
Prices
Speech models on the API. Billed prices, as charged.
| Name | Maker | Price | Detail |
|---|---|---|---|
| openai/whisper-large-v3-turbo | OpenAI | $0.00021 per minute | Whisper's faster decoder |
| openai/whisper-large-v3 | OpenAI | $0.000473 per minute | The full Whisper model |
| mistralai/Voxtral-Small-24B-2507 | Mistral AI | $0.00315 per minute | Mistral's speech model, multilingual |
Why switch
Why teams move to Runcrate. One account for models and GPUs.
The OpenAI audio endpoint
Existing client.audio.transcriptions.create() calls work with a different base URL and key: json, verbose_json with timed segments, text, srt and vtt.
Billed per second of audio
Each file is charged for its length at the per-minute price, with no per-request minimum, from the same prepaid balance as every other model.
Understanding with any chat model
Send the transcript to a chat completion for a summary, topics or sentiment, and pick the model for that step: DeepSeek, Qwen, Llama or a frontier model.
GPUs for the rest
For diarization or a model the API does not serve, start a GPU instance on the same account and run an open model such as pyannote yourself.
Get started
Try it in one request.
Your agent writes and runs it. The same key works for every model and every GPU.
FAQ