German speech → usable text
German transcription. Find your API.
Language support is the first filter, not the finish line. Shortlist German-capable APIs, estimate your workload and check independent German recognition results below.
Developer APIs, not meeting-assistant subscriptions.
Ranked for your volume
Transcription monthly cost by provider
Costs update as you type.
Equal-sized requests; one audio channel.
Whisper at Cloudflare Workers AI
17 priced options · German confirmed · base transcription
17 options with a full price · 6 more need a closer look (below)
| Workflow / evidence | ||||
|---|---|---|---|---|
| WhisperCloudflare Workers AILowest estimate | $2.70 | Not benchmarked | $0.027base rate | Recorded filesSources & billing notesWorkers AI duration estimate; free allocation and account charges are not deducted. Price snapshot: 2026-10-31. Tracked dataset. Official pricing ↗ · Feature evidence ↗ |
| Whisper Large v3 TurboCloudflare Workers AI | $3.06 | 3.8%FLEURS · lower is better | $0.031base rate | Recorded filesSources & billing notesWorkers AI duration estimate; dataset rounds the provider's per-minute rate. Free allocation is not deducted. Price snapshot: 2026-10-31. Tracked dataset. Official pricing ↗ · Feature evidence ↗Accuracy: Open ASR Leaderboard result for |
| Whisper Large v3 TurboGroq | $4.00 | 3.8%FLEURS · lower is better | $0.04base rate | Recorded filesSources & billing notes10-second minimum billed duration per request. Price snapshot: 2026-10-31. Tracked dataset. Official pricing ↗ · Feature evidence ↗Accuracy: Open ASR Leaderboard result for |
| Whisper Large v3Groq | $11.10 | 3.3%FLEURS · lower is better | $0.111base rate | Recorded filesSources & billing notes10-second minimum billed duration per request. Price snapshot: 2026-10-31. Tracked dataset. Official pricing ↗ · Feature evidence ↗Accuracy: Open ASR Leaderboard result for |
| Melia 1Speechmatics | $12.00 | Not benchmarked | $0.12base rate | Recorded filesSources & billing notesPre-recorded pay-as-you-go rate. Multilingual model; speaker diarization included. Price snapshot: 2026-10-17. Sourced preview; confirm the provider quote. Official pricing ↗ · Feature evidence ↗ |
| Universal-2AssemblyAI | $15.00 | Not benchmarked | $0.15base rate | Recorded filesSources & billing notesPay-as-you-go pre-recorded rate per hour of audio submitted. Speaker labels +$0.02/hr. Price snapshot: 2026-10-17. Sourced preview; confirm the provider quote. Official pricing ↗ · Feature evidence ↗ |
| Azure Speech batch transcriptionMicrosoft Azure | $18.00 | Not benchmarked | $0.18base rate | Recorded filesSources & billing notesStandard batch rate, billed per second. Real-time transcription costs $1 per hour. Price snapshot: 2026-10-17. Sourced preview; confirm the provider quote. Official pricing ↗ · Feature evidence ↗ |
| Speech-to-Text V2 dynamic batchGoogle Cloud | $18.00 | Not benchmarked | $0.18base rate | Recorded filesSources & billing notesDynamic batch recognition for standard models: much cheaper, results are not immediate. Check the latency terms. Price snapshot: 2026-10-17. Sourced preview; confirm the provider quote. Official pricing ↗ · Feature evidence ↗ |
| Voxtral Mini Transcribe 2Mistral AI | $18.00 | Not benchmarked | $0.18base rate | Recorded filesSources & billing notesBase transcription estimate; account billing rules can affect the invoice. Price snapshot: 2026-10-31. Tracked dataset. Official pricing ↗ · Feature evidence ↗ |
| Universal-3.5 ProAssemblyAI | $21.00 | Not benchmarked | $0.21base rate | Recorded filesSources & billing notesPay-as-you-go pre-recorded rate per hour of audio submitted. Speaker labels +$0.02/hr. Price snapshot: 2026-10-17. Sourced preview; confirm the provider quote. Official pricing ↗ · Feature evidence ↗ |
Price not fully listed 1
These match your requirements, but the base price or the speaker-label charge is not published in a form we can calculate.
Might fit, not documented 5
The provider does not document at least one of your requirements. That does not mean it is unsupported: test it or ask the provider.
Accuracy is the word error rate (WER) the open Open ASR Leaderboard measured for the exact same model (English average, or FLEURS for the selected language); it is not a test of your audio, and rows without an exact match say so. Costs are API usage estimates, not all-in account bills. Includes documented request minimums and known speaker-label charges when required. Excludes taxes, plan fees, credits, retries, storage and other extras. Incomplete language evidence is kept out of a confirmed shortlist. Test representative audio before integrating.
Price is only half the decision
German recognition: a separate benchmark view
These are existing Open ASR Leaderboard results on the German FLEURS dataset. Word error rate (WER) is lower-is-better. This is not an AISelect test and not a ranking of API hosts. Different versions, domains, punctuation rules and audio conditions can change the outcome.
| Evaluated model / endpoint | WER (%) ↓ |
|---|---|
| elevenlabs/scribe_v2 | 2.30% |
| assembly/universal-3-pro | 2.32% |
| mistralai/Voxtral-Small-24B-2507 | 2.59% |
| openai/whisper-large-v3 | 3.27% |
| Qwen/Qwen3-ASR-1.7B | 3.33% |
| nvidia/canary-1b-v2 | 3.47% |
| speechmatics/enhanced | 3.69% |
| facebook/omniASR-LLM-7B-v2 | 3.73% |
| microsoft/Phi-4-multimodal-instruct | 3.77% |
| openai/whisper-large-v3-turbo | 3.78% |
| mistralai/Voxtral-Mini-3B-2507 | 3.82% |
| facebook/omniASR-LLM-7B | 4.04% |
| CohereLabs/cohere-transcribe-03-2026 | 4.16% |
| nvidia/parakeet-tdt-0.6b-v3 | 4.26% |
| facebook/omniASR-LLM-3B-v2 | 4.84% |
| facebook/omniASR-CTC-7B-v2 | 4.96% |
| mistralai/Voxtral-Mini-4B-Realtime-2602 | 5.04% |
| facebook/omniASR-CTC-3B-v2 | 5.13% |
| facebook/omniASR-CTC-7B | 5.33% |
| Qwen/Qwen3-ASR-0.6B | 5.49% |
| facebook/omniASR-LLM-3B | 5.75% |
| facebook/omniASR-LLM-1B | 6.11% |
| facebook/omniASR-CTC-3B | 6.14% |
| facebook/omniASR-LLM-1B-v2 | 7.17% |
| facebook/omniASR-CTC-1B-v2 | 7.44% |
| facebook/omniASR-CTC-1B | 8.05% |
| facebook/omniASR-LLM-300M-v2 | 8.07% |
| facebook/omniASR-LLM-300M | 9.16% |
| facebook/omniASR-CTC-300M-v2 | 14.59% |
| facebook/omniASR-CTC-300M | 14.94% |
Source file last changed 2026-03-26; retrieved 2026-10-16. Retrieval is not a new evaluation. Source CSV revision ↗ · Methodology and Apache-2.0 source ↗. The listed Voxtral Mini 2507 score does not describe Transcribe 2. Whisper model scores do not certify Groq or Cloudflare hosting quality.
Your 10-minute German test
Use representative audio: one clean recording, one noisy conversation and your actual vocabulary. Check names, numbers and dialect. Compare correction time as well as WER; a slightly higher API bill can be cheaper than fixing the transcript.