Rev AI
Low-cost batch speech-to-text with structured JSON, word timestamps, and speaker labels, but mixed accuracy on crosstalk and code-switching.
Good value for batch transcription, but not a safe default for hard multilingual audio.
- You need a low-cost batch STT API that returns structured JSON with word-level timestamps, confidence values, and speaker labels.
- You can process long audio asynchronously and are fine auditing difficult clips manually.
- You want measured throughput around $0.1998/audio-hour rather than a higher-cost transcription service.
- You need verified speaker attribution on overlap; the crosstalk run over-segmented speakers and DER was not measured.
Our take
Rev AI consistently returned structured transcript JSON with word-level timestamps, confidence values, and speaker labels, and the benchmarked Reverb Transcription tier came out to $0.1998/audio-hour. The tradeoff is accuracy: it scored 9.79% WER on medical jargon, but 28.33% on overlapping speech and 25.16% on bilingual code-switching, with only 8.8% Spanish token recall.
In-Depth Review
Our detailed analysis of Rev AI — features, performance, and real-world testing.
Feature-by-Feature Breakdown
Asynchronous Batch Transcription▾
Feature tested: Asynchronous Batch Transcription
Result: Partial
Expected behavior: Rev AI submits long-form audio as asynchronous speech-to-text jobs, then lets you poll and fetch the transcript once processing completes. The cards exercise this POST → poll → fetch flow on multiple long-audio files, including crosstalk.wav, medical_terms.mp3, and mix_language.mp3.
Test case: Audio file → Text/code file
Input type: Audio file
Input used: Input artifact (Audio file): INPUT: Overlapping Speech / Crosstalk — crosstalk.wav, 65.39 MB, 2142.709s, overlapping meeting audio. — crosstalk.wav
Observed output: Output artifact (Text/code file): Returned transcript JSON for the crosstalk sample; the scored run completed successfully and recorded 28.33% WER, 102.7s latency, and 6 detected speaker labels. — raw-response.json
Input artifact: Input artifact (Audio file): INPUT: Overlapping Speech / Crosstalk — crosstalk.wav, 65.39 MB, 2142.709s, overlapping meeting audio. — crosstalk.wav
Output artifact: Output artifact (Text/code file): Returned transcript JSON for the crosstalk sample; the scored run completed successfully and recorded 28.33% WER, 102.7s latency, and 6 detected speaker labels. — raw-response.json
What changed: Audio file transformed into Text/code file
Test case: Audio file → Text/code file
Input type: Audio file
Input used: Input artifact (Audio file): INPUT: Medical Jargon — medical_terms.mp3, 8.58 MB, 1123.944s, single-speaker anatomical narration. — medical_terms.mp3
Observed output: Output artifact (Text/code file): Returned transcript JSON for the medical-jargon sample; the scored run completed successfully and recorded 9.79% WER, 78.32s latency, and a single detected speaker label. — raw-response-2.json
Input artifact: Input artifact (Audio file): INPUT: Medical Jargon — medical_terms.mp3, 8.58 MB, 1123.944s, single-speaker anatomical narration. — medical_terms.mp3
Output artifact: Output artifact (Text/code file): Returned transcript JSON for the medical-jargon sample; the scored run completed successfully and recorded 9.79% WER, 78.32s latency, and a single detected speaker label. — raw-response-2.json
What changed: Audio file transformed into Text/code file
Test case: Audio file → Text/code file
Input type: Audio file
Input used: Input artifact (Audio file): INPUT: Bilingual Code-Switching — mix_language.mp3, 22.19 MB, 1938.495s, mixed Spanish-English conversation. — mix_language.mp3
Observed output: Output artifact (Text/code file): Returned transcript JSON for the mixed-language sample; the scored run completed successfully and recorded 25.16% WER, 99.79s latency, and 3 detected speaker labels. — raw-response-3.json
Input artifact: Input artifact (Audio file): INPUT: Bilingual Code-Switching — mix_language.mp3, 22.19 MB, 1938.495s, mixed Spanish-English conversation. — mix_language.mp3
Output artifact: Output artifact (Text/code file): Returned transcript JSON for the mixed-language sample; the scored run completed successfully and recorded 25.16% WER, 99.79s latency, and 3 detected speaker labels. — raw-response-3.json
What changed: Audio file transformed into Text/code file
Why it matters / Conclusion: The async integration pattern worked cleanly across all three files, returning transcript JSON each time; the benchmark's quality differences came from the audio difficulty, not from request handling failures.
Rev AI submits long-form audio as asynchronous speech-to-text jobs, then lets you poll and fetch the transcript once processing completes. The cards exercise this POST → poll → fetch flow on multiple long-audio files, including crosstalk.wav, medical_terms.mp3, and mix_language.mp3.
Structured Transcript Output with MetadataThe API consistently exposed rich transcript metadata for downstream parsing.▾
Feature tested: Structured Transcript Output with Metadata
Result: Partial
Verdict: The API consistently exposed rich transcript metadata for downstream parsing.
Expected behavior: Rev AI returns transcript JSON with rich metadata such as word-level timestamps, confidence values, punctuation, timed word tokens, monologues, and speaker-label fields. The cards exercise this structured payload on repeated runs and a word-level metadata variant.
Test case: Audio file → Text/code file
Input type: Audio file
Input used: Input artifact (Audio file): INPUT: Overlapping Speech / Crosstalk — transcript payload returned for a 4-speaker meeting with crosstalk. — crosstalk.wav
Observed output: Output artifact (Text/code file): The raw JSON transcript included word-level timestamps and confidence values, along with punctuation and speaker fields. — raw-response.json
Input artifact: Input artifact (Audio file): INPUT: Overlapping Speech / Crosstalk — transcript payload returned for a 4-speaker meeting with crosstalk. — crosstalk.wav
Output artifact: Output artifact (Text/code file): The raw JSON transcript included word-level timestamps and confidence values, along with punctuation and speaker fields. — raw-response.json
What changed: Audio file transformed into Text/code file
Test case: Audio file → Text/code file
Input type: Audio file
Input used: Input artifact (Audio file): INPUT: Medical Jargon — transcript payload returned for dense anatomical narration. — medical_terms.mp3
Observed output: Output artifact (Text/code file): The raw JSON transcript included word-level timestamps and confidence values, and the run reported 2,779 timed tokens with payload depth 3/3. — raw-response-2.json
Input artifact: Input artifact (Audio file): INPUT: Medical Jargon — transcript payload returned for dense anatomical narration. — medical_terms.mp3
Output artifact: Output artifact (Text/code file): The raw JSON transcript included word-level timestamps and confidence values, and the run reported 2,779 timed tokens with payload depth 3/3. — raw-response-2.json
What changed: Audio file transformed into Text/code file
Test case: Audio file → Text/code file
Input type: Audio file
Input used: Input artifact (Audio file): INPUT: Bilingual Code-Switching — transcript payload returned for mixed Spanish-English speech. — mix_language.mp3
Observed output: Output artifact (Text/code file): The raw JSON transcript included word-level timestamps and confidence values, with 5,855 timed tokens and payload depth 3/3. — raw-response-3.json
Input artifact: Input artifact (Audio file): INPUT: Bilingual Code-Switching — transcript payload returned for mixed Spanish-English speech. — mix_language.mp3
Output artifact: Output artifact (Text/code file): The raw JSON transcript included word-level timestamps and confidence values, with 5,855 timed tokens and payload depth 3/3. — raw-response-3.json
What changed: Audio file transformed into Text/code file
Why it matters / Conclusion: The structured payload is consistent and integration-friendly across all three runs.
Rev AI returns transcript JSON with rich metadata such as word-level timestamps, confidence values, punctuation, timed word tokens, monologues, and speaker-label fields. The cards exercise this structured payload on repeated runs and a word-level metadata variant.
Speaker DiarizationSpeaker labels are always present, but diarization correctness was not verified.▾
Feature tested: Speaker Diarization
Result: Partial
Verdict: Speaker labels are always present, but diarization correctness was not verified.
Expected behavior: Rev AI emits speaker labels in transcript output so multi-speaker audio can be separated by speaker. The cards cover crosstalk and overlapping-meeting samples where labels were present, though segmentation quality varied.
Test case: Audio file → Text/code file
Input type: Audio file
Input used: Input artifact (Audio file): INPUT: Overlapping Speech / Crosstalk — 4-participant meeting audio with overlap. — crosstalk.wav
Observed output: Output artifact (Text/code file): The crosstalk payload exposed speaker labels, and the run metrics reported 6 detected labels on a 4-participant recording, indicating over-segmentation. — raw-response.json
Input artifact: Input artifact (Audio file): INPUT: Overlapping Speech / Crosstalk — 4-participant meeting audio with overlap. — crosstalk.wav
Output artifact: Output artifact (Text/code file): The crosstalk payload exposed speaker labels, and the run metrics reported 6 detected labels on a 4-participant recording, indicating over-segmentation. — raw-response.json
What changed: Audio file transformed into Text/code file
Test case: Audio file → Text/code file
Input type: Audio file
Input used: Input artifact (Audio file): INPUT: Medical Jargon — single-speaker narration. — medical_terms.mp3
Observed output: Output artifact (Text/code file): The medical-jargon payload exposed a single speaker label in the transcript JSON. — raw-response-2.json
Input artifact: Input artifact (Audio file): INPUT: Medical Jargon — single-speaker narration. — medical_terms.mp3
Output artifact: Output artifact (Text/code file): The medical-jargon payload exposed a single speaker label in the transcript JSON. — raw-response-2.json
What changed: Audio file transformed into Text/code file
Test case: Audio file → Text/code file
Input type: Audio file
Input used: Input artifact (Audio file): INPUT: Bilingual Code-Switching — mixed-language conversation with multiple speakers. — mix_language.mp3
Observed output: Output artifact (Text/code file): The bilingual payload exposed speaker labels and the run metrics reported 3 detected speakers. — raw-response-3.json
Input artifact: Input artifact (Audio file): INPUT: Bilingual Code-Switching — mixed-language conversation with multiple speakers. — mix_language.mp3
Output artifact: Output artifact (Text/code file): The bilingual payload exposed speaker labels and the run metrics reported 3 detected speakers. — raw-response-3.json
What changed: Audio file transformed into Text/code file
Why it matters / Conclusion: Speaker-label presence is reliable, but crosstalk was over-segmented and attribution quality remains unverified.
Rev AI emits speaker labels in transcript output so multi-speaker audio can be separated by speaker. The cards cover crosstalk and overlapping-meeting samples where labels were present, though segmentation quality varied.
Multilingual and Code-Switching TranscriptionThe benchmarked English tier handled Spanish-English code-switching poorly.▾
Feature tested: Multilingual and Code-Switching Transcription
Result: Failed
Verdict: The benchmarked English tier handled Spanish-English code-switching poorly.
Expected behavior: Rev AI attempts transcription on mixed-language audio, including English/Spanish code-switching examples. The cards show the capability on bilingual samples, but also note weak Spanish recall in the tested configuration.
Test case: Audio file → Text/code file
Input type: Audio file
Input used: Input artifact (Audio file): INPUT: Bilingual Code-Switching — mix_language.mp3, a Spanish-English conversation used to test mixed-language handling. — mix_language.mp3
Observed output: Output artifact (Text/code file): The transcript was produced, but the benchmarked English tier scored 25.16% WER and only 8.8% Spanish token recall, missing tokens such as ahora and turning mi entonces ahora into English words. — raw-response-3.json
Input artifact: Input artifact (Audio file): INPUT: Bilingual Code-Switching — mix_language.mp3, a Spanish-English conversation used to test mixed-language handling. — mix_language.mp3
Output artifact: Output artifact (Text/code file): The transcript was produced, but the benchmarked English tier scored 25.16% WER and only 8.8% Spanish token recall, missing tokens such as ahora and turning mi entonces ahora into English words. — raw-response-3.json
What changed: Audio file transformed into Text/code file
Why it matters / Conclusion: Do not treat the benchmarked English configuration as a safe multilingual default; Spanish-heavy code-switching was mostly mistranscribed.
Rev AI attempts transcription on mixed-language audio, including English/Spanish code-switching examples. The cards show the capability on bilingual samples, but also note weak Spanish recall in the tested configuration.
Official pricing
The benchmarked Reverb Transcription tier is $0.20/audio-hour, and Rev AI also offers a free-credit entry tier plus a separate multilingual tier.
Source: https://www.rev.ai/pricing, accessed 2026-08-14. The benchmark used Reverb Transcription.
Banner Preview
How the embed badge will look on your site
Embed HTML
Copy this code to your website source
Quick Integration Guide
- 1Copy the HTML code block above.
- 2Paste it into your site's HTML or CMS editor.
- 3Banner appears instantly on your page.
- 4Links back to your tool profile here.
Similar Tools
Discover more AI tools like Rev AI to enhance your workflow.
Comments (0)
Need a custom AI solution for this use case?
If you are looking to build a custom speech-to-text, audio transcription, or transcription system for your business or internal workflow, email us at contact@futuresmart.ai.
Found something inaccurate or missing? We try to keep our AI research accurate and useful. If you found outdated information, an issue, or have a suggestion, email us at collaborate@aidemos.com.