OpenAI Launches GPT Transcription Models for Streaming and Batch Audio, Cutting Price to $4.50 Per 1,000 Minutes
OpenAI has launched two new speech-to-text models in its API, GPT-Transcribe and GPT-Live-Transcribe, designed for batch processing and low-latency streaming, respectively. Both models integrate free-form text prompts, keywords, and language hints to improve semantic accuracy, with the live version reaching a 44.6% semantic accuracy score when context is provided.
The batch model is priced at $4.50 per 1,000 minutes of audio, down 25% from the prior model. On the 22-language Common Voice benchmark, GPT-Transcribe posted a 19.27% error rate, beating the older Whisper-1 base, while the live version hit a 9.60% error rate compared to 11.65% for GPT-Realtime-Whisper-1 on the Real-World Audio Recording benchmark. OpenAI plans to publish performance metrics for the streaming model on its public leaderboards later this month.
From the sources (15 posts)
@openaidevsWe're introducing two new transcription models in the API: • GPT-Live-Transcribe: built for low-latency live transcription. • GPT-Transcribe: optimized for asynchronous transcription of completed audio files and batch workloads. Both mode
@openaidevsBuilders can improve live audio transcription by providing: • Free-form context about the recording • Keywords for names and domain-specific terms • Expected input languages • Earlier transcribed turns as context On our new Context Aware
@openaidevsGPT-Transcribe also improved on the Context Aware ASR benchmark, with semantic accuracy increasing from 41.6% without free-form context to 45.2% with it. Across 22 languages on Common Voice, GPT-Transcribe achieved a 19.27% transcription e
@openaidevs
@openaidevsWe're introducing two new transcription models in the API: • GPT-Live-Transcribe: built for low-latency live transcription. • GPT-Transcribe: optimized for asynchronous transcription of completed audio files and batch workloads. Both mode
@openaidevsBuilders can improve live audio transcription by providing: • Free-form context about the recording • Keywords for names and domain-specific terms • Expected input languages • Earlier transcribed turns as context On our new Context Aware
@openaidevsGPT-Transcribe also improved on the Context Aware ASR benchmark, with semantic accuracy increasing from 41.6% without free-form context to 45.2% with it. Across 22 languages on Common Voice, GPT-Transcribe achieved a 19.27% transcription e
@openaidevs
@thezachmuellerRT @OpenAIDevs: We're introducing two new transcription models in the API: • GPT-Live-Transcribe: built for low-latency live transcription…
@jubertiNew OpenAI transcription models! gpt-transcribe is our new offline ASR, replacing gpt-4o-transcribe, and gpt-live-transcribe is our new realtime ASR, replacing gpt-realtime-whisper. These models have better WER and support better hinting (e
@testingcatalogOPENAI 🔥: New GPT-Live-Transcribe and GPT-Transcribe models are now available on APIs! > GPT-Live-Transcribe - low-latency live transcription. > GPT-Transcribe - asynchronous transcription of completed audio files and batch workloa
@reach_vbTwo new transcription models are now available in the API! > GPT Live Transcribe for low-latency live transcription > GPT Transcribe for completed audio and batch workloads The coolest part is how well they can use context - you can pass
@wesrothOpenAI just released two new transcription models. One listens live. The other processes completed recordings at scale. GPT-Live-Transcribe is built for real-time captions, calls, and streaming audio. GPT-Transcribe is optimized for uplo
@artificialanlysOpenAI has released GPT Transcribe: a Speech to Text model scoring 3.31% on AA-WER (#9), improving 0.7 p.p. over its predecessor GPT-4o Transcribe while lowering price 25% to $4.50 per 1,000 minutes of audio GPT Transcribe is OpenAI's late
@artificialanlysGPT Transcribe is available at $4.50 per 1,000 minutes of audio ($0.0045/min) via the OpenAI API Platform, 25% cheaper than its predecessor GPT-4o Transcribe ($6.00 per 1,000 minutes), though GPT-4o Mini Transcribe remains available at $3.0