The AI transcription race in 2026 is no longer about "can it hear the words." It is about four things: how accurate, how good is Chinese, can it produce minutes and subtitles, and how much does it cost. Open any tool list and there are easily twenty products, all jostling for a name: Whisper, Otter, Tingwu, Miaojí, Notta… newcomers can't tell them apart. This piece picks five of the most representative and compares them side by side, giving you a per-scenario verdict instead of a vanity ranking.
One thing up front: every comparison below is a representative comparison based on each tool's official docs, pricing pages, and public reviews, not a benchmark I ran myself. Prices and features move, and by the time you read this something will have updated, so treat the official latest word as the source of truth, not this article. Because some of the figures here come from training data rather than real-time verification, every concrete number is marked "approx" or "verify on the official site," so please double-check on the vendor's site before you buy.
One more premise: the five tools here cover three deployment shapes. OpenAI's Whisper is a hybrid API+open-source product; Otter.ai and Notta are overseas pure SaaS; Tongyi Tingwu and Feishu Miaojí are domestic Chinese SaaS. Five directions, three playbooks, mapping the 2026 transcription landscape.
1. Why you need to compare transcription tools side by side in 2026
The reason the field is crowded is that the category filled three long-standing gaps in 2025-2026. First, multilingual coverage. Whisper's 50+ languages, Notta's 50+ languages, and Tingwu's Chinese-first tuning basically solved "can it understand my language." Second, minutes capability. We moved from raw transcription to meeting minutes, action items, chapter summaries, and AI Q&A; the transcript went from "a pile of words" to "something ready to use." Third, ecosystem integration. Otter hooks into Zoom/Teams/Meet, Miaojí lives inside Feishu meetings, Tingwu connects to DingTalk, and tools are no longer islands.
But even as the gaps closed, the five tools each bet on a different direction, and none is the all-rounder. Whisper bets on accuracy and open source, Otter on English meetings and real-time minutes, Tingwu on Chinese and the Alibaba ecosystem, Miaojí on zero-friction integration inside Feishu, and Notta on multilingual coverage and cross-platform. Hold those bets in your head and you will not need a ranking; you will just look at your own job.
The five questions to ask yourself before choosing: are your meetings mostly Chinese or English, do you want raw transcripts or finished minutes and subtitles, are you an individual or a team, can your data leave the country, and are you a developer who wants an API or a beginner who wants a UI? Nail those five and the choice below is basically locked.
2. The five contenders (data as of 2026-07-30)
Here is the lineup. Pricing is from each vendor's site or pricing page; open-source status and model versions are from official sources. Figures marked "approx" were not real-time verified; please confirm on the official site.
| Tool | Vendor | Latest model | Open source | Free tier | Starting price (approx, verify on site) | Deployment |
|---|---|---|---|---|---|---|
| Whisper | OpenAI | whisper-1 / large-v3 | Yes (open weights) | None on API; free if self-hosted | $0.006/min (approx) | API/local |
| Otter.ai | AISense | Otter (2026) | No | 300 min/month | $16.99/mo (approx) | Web/app |
| Tongyi Tingwu | Alibaba Cloud | Tingwu (Qwen-based) | No | Free hours (approx 5-10h/mo, verify on site) | ~¥99/mo | Web/app |
| Feishu Miaojí | ByteDance | Miaojí (Doubao-based) | No | Included with Feishu account | Bundled in Feishu paid plans (~¥60/user/mo) | Web/desktop |
| Notta | Notta Inc. | Notta (2026) | No | 120 min/month | $13.99/mo (approx, lower if annual) | Web/app |
Two details to call out. First, on free tiers, Otter (300 min), Tingwu (free hours), Miaojí (Feishu quota), and Notta (120 min) all give real free quota; Whisper's "free" means you run the open-source version yourself. Second, on open source, only Whisper has open weights (GitHub repo openai/whisper); the other four are all closed SaaS. These two points set your starting move: if your data can't leave your machine, run Whisper locally; if you want it to just work, you have to accept SaaS.
3. Side by side: six dimensions laid flat
Lay the official descriptions flat and line them up across six dimensions. This table is a representative comparison based on official docs and public descriptions, not a hands-on benchmark.
| Dimension | Whisper | Otter.ai | Tongyi Tingwu | Feishu Miaojí | Notta |
|---|---|---|---|---|---|
| Accuracy (English) | Top tier | Top tier | Upper-mid | Upper-mid | Top tier |
| Accuracy (Chinese) | Upper-mid | Weak-to-mid (English-first) | Strongest (native Chinese) | Strongest (native Chinese) | Upper-mid |
| Multilingual | 50+ languages | Mainly English, weak multilingual | Chinese and English mainly, others weak | Chinese, English, Japanese mainly | 50+ languages |
| Real-time transcription | No (API/local, mostly non-streaming) | Yes | Yes | Yes (in-meeting) | Yes |
| Minutes / summary | No (raw transcript, build your own) | Yes (minutes, action items) | Yes (chapter summaries, Q&A) | Yes (AI assistant, todos) | Yes (summary, chapters) |
| Subtitle export | Yes (SRT/VTT, DIY) | Limited | Yes | Limited | Yes |
| Ecosystem integration | API/dev frameworks | Zoom/Teams/Meet/Slack | DingTalk/Alibaba Cloud | Full Feishu suite (tight bind) | Zoom/Meet/Teams/Notion |
| Best for | Developers/local/privacy | English meeting teams | Chinese meeting teams | Feishu users | Multilingual/cross-platform |
Mind the apples-to-oranges: accuracy can't be measured on one numeric scale, so the table is qualitative, based on representative public reviews. Whisper and Otter are the acknowledged top tier on English, but Whisper has no built-in real-time streaming or minutes; you assemble that with the API yourself. On Chinese, Tingwu and Miaojí are the home team, natively trained by Alibaba and ByteDance; the other three need English to perform. On ecosystem integration Miaojí is special: it is tightly bound to Feishu, zero-friction for Feishu users but almost unusable outside Feishu.
4. One by one: each tool's best range
Whisper: the open-source king, the only real answer for local deployment and developer integration
Whisper is the speech recognition model OpenAI open-sourced in 2022; into 2026 it remains the de facto standard in open-source transcription. It splits in two: the API via OpenAI's official endpoint, around $0.006 per minute (figure per OpenAI's site); and the open-source form via the openai/whisper GitHub repo, weights public, which you can pull down, run locally, even fine-tune. It supports 50+ languages for transcription and translation to English, with acknowledged top-tier accuracy on English and upper-mid on Chinese.
Its bet is "accuracy plus open source." Open source means three things: your data never leaves your machine (a hard requirement for privacy-sensitive work), you can fine-tune domain vocabulary, and you can run it free forever (you pay for electricity and the GPU). For teams wiring transcription into a product, batch-processing audio, or with hard privacy needs, Whisper has almost no substitute. The cost is the barrier: it is just a model and an API, no UI, no real-time streaming minutes, no meeting integration, you assemble the product yourself. Running it locally also needs a big-VRAM GPU (large-v3 starts around 10GB), and tuning has a learning curve.
Best for: developers, teams that need local or batch transcription, privacy-sensitive scenarios, and domain applications that need custom vocabularies. Beginners and anyone who wants ready-made meeting minutes should look elsewhere.
Otter.ai: the benchmark for English meetings, strongest on real-time minutes
Otter.ai is the best-known meeting transcription SaaS overseas; its bet is "English meetings plus real-time minutes." Its strongest scene is English Zoom/Teams/Meet meetings: real-time transcription, automatic speaker identification, automatic post-meeting minutes and action items, an end-to-end experience that has almost no rival in English. Free tier 300 minutes/month, 30 minutes per session; Pro about $16.99/month, 1200 minutes, 90 minutes per session; Business about $35/user/month, 6000 minutes; Enterprise custom.
Its bet is "English meetings." If your team is English-first and wants to skip the post-meeting minutes grind, Otter is the easy choice. The tradeoffs: weak Chinese support (multilingual is not its main field), per-user pricing gets pricey for teams, and data sits on its servers (a concern for privacy-sensitive work). Its Chinese accuracy and minutes quality are clearly worse than English; domestic teams should not pick it for Chinese meetings.
Best for: English-first overseas teams, PMs and sales who want real-time minutes and action items, and teams that can accept per-user pricing. For Chinese-first meetings, look at Tingwu or Miaojí.
Tongyi Tingwu: best range for Chinese meetings, native to the Alibaba ecosystem
Tongyi Tingwu is a transcription product from Alibaba Cloud; its bet is "Chinese plus the Alibaba ecosystem." Its strongest scene is Chinese meetings: Chinese accuracy is among the strongest of the five (tied with Miaojí), with chapter summaries, AI Q&A, and action-item extraction, deeply integrated with DingTalk and Alibaba Cloud Drive. There is a free quota (around 5-10 hours/month, per the official site), paid starting around ¥99/month, with pay-per-use also available.
Its bet is "native Chinese." Alibaba-trained and Chinese-tuned, it clearly outperforms English tools on colloquial Chinese, specialized terminology, and names and places. For teams that want Chinese meeting minutes, DingTalk ecosystem integration, and data kept domestic, Tingwu is the easiest choice. The tradeoffs: English and multilingual are not its strengths, the interface is Chinese-first and unfriendly to overseas users, and it binds to the Alibaba stack.
Best for: Chinese-first domestic teams, teams on DingTalk or Alibaba Cloud, and domestic enterprises with data-residency requirements. For English-first meetings, look at Otter.
Feishu Miaojí: zero-friction integration inside the Feishu suite
Feishu Miaojí is the transcription feature inside ByteDance's Feishu (Lark) suite; its bet is "zero friction in Feishu." Its strongest scene is Feishu meetings: transcription starts automatically, the minutes appear in Feishu Docs after the meeting, the AI assistant answers questions from the minutes, and todos are generated automatically. Chinese accuracy ties with Tingwu for the strongest. The Feishu account comes with a Miaojí quota; higher quota is bundled in Feishu paid plans (around ¥60/user/month and up, per the official site).
Its bet is "integration plus Chinese." For teams already on Feishu, Miaojí is almost zero friction: no second tool to buy, no exporting back and forth, minutes and todos natively linked to Feishu Docs. The tradeoffs: tightly bound to Feishu, almost unusable outside Feishu, no independent per-minute pricing, and zero value for non-Feishu users. Its worth depends entirely on whether you are a Feishu user.
Best for: Feishu users, and teams that want meeting minutes deeply linked to Feishu Docs. If you are not on Feishu, Tingwu or Otter fit better.
Notta: the all-rounder of multilingual and cross-platform
Notta is a multilingual transcription SaaS from a Japanese team; its bet is "multilingual plus cross-platform." It supports 50+ languages, with English and Japanese as strengths and Chinese in the upper-mid. Free tier 120 minutes/month, Pro about $13.99/month (lower if annual), Business about $23.99/user/month, Enterprise custom. Features include real-time transcription, file import, translation, summaries, and chapter division, with Zoom/Meet/Teams/Notion integrations.
Its bet is "multilingual coverage." If you do cross-language transcription — say an English meeting plus a Japanese interview, or a French podcast with Chinese subtitles — Notta has the widest coverage. The tradeoffs: Chinese accuracy trails Tingwu and Miaojí, English trails the top tier of Otter, minutes quality is competent but not the strongest. It is more a multilingual Swiss army knife, not the best at any one language but workable at all.
Best for: cross-language transcription needs, multilingual podcast and subtitle creators, and overseas individuals and small teams that want Zoom/Notion integration. For pure Chinese or pure English scenes, the previous four are more specialized.
5. A decision tree: three scenarios to slot yourself into
Don't pick by hype; pick by the job. Here are decision paths for three high-frequency scenarios.
Scenario 1: meeting minutes. After a meeting you want minutes, action items, todos. If your meetings are mostly Chinese — pick Tongyi Tingwu or Feishu Miaojí. Miaojí for Feishu users (zero friction), Tingwu for non-Feishu domestic teams. If your meetings are mostly English — pick Otter.ai; real-time minutes and action items are its home turf. If you are a developer building your own meeting-minutes pipeline — pick the Whisper API and wire in an LLM for summaries; flexible but more work.
Scenario 2: podcast transcription. You want to turn podcast audio into text for publishing, with chapter summaries. For Chinese podcasts — pick Tongyi Tingwu; Chinese accuracy is the steadiest, chapter summaries ready to go. For English podcasts — pick Notta or Otter; Notta is better for file import and multilingual coverage, Otter for real-time live scenarios. For batch processing with sensitive data — pick Whisper local; open source, runs locally, batch-friendly, fine-tunable for domain vocab.
Scenario 3: subtitle production. You want SRT/VTT subtitles for video. For Chinese subtitles — pick Tongyi Tingwu or Feishu Miaojí; accurate Chinese timecodes. For multilingual subtitles — pick Notta or the Whisper API; Notta has built-in multilingual export, Whisper is flexible but you handle the timecode format yourself. For developers building subtitle tools — pick the Whisper API; open, integrable, customizable output formats.
A few common combos. One, Whisper as the transcription layer plus Qwen or GPT as the LLM for minutes; developers assemble their own meeting system via API. Two, Feishu Miaojí for internal meetings plus Tingwu for external interviews, divided by scene. Three, Otter for English meetings plus Notta for cross-language podcasts, a bilingual combo for overseas teams. Transcription tools are not a single-choice question; pairing by scenario beats betting on one all-rounder.
6. Four pitfalls: data residency, free-tier caps, minutes quality, and pricing units
First, data residency is the biggest hidden trap. Whisper API, Otter, and Notta run their servers overseas; uploading your meeting audio means data leaves the country. For meetings that involve customer privacy, trade secrets, or personal information, do not upload to overseas SaaS for convenience. In those cases pick a domestic SaaS (Tingwu, Miaojí) or run Whisper locally, so data never leaves.
Second, free-tier caps are hidden traps. Otter's free 300 minutes cap each session at 30 minutes, so a one-hour meeting only records half. Notta's free 120 minutes is basically enough to try. Tingwu and Miaojí's free quotas are also limited. Free is fine for personal tinkering and light use; once you scale, the caps bite. Switch to paid at that point; don't pinch pennies stuck on free.
Third, minutes quality matters more than raw accuracy. The transcript goes from "a pile of words" to "something ready to use" through minutes capability. Whisper transcribes accurately but does not do minutes; you wire in an LLM yourself. Otter, Tingwu, Miaojí, and Notta all ship minutes, but quality differs a lot — Otter is strongest on English minutes, Tingwu and Miaojí on Chinese, Notta on cross-language. Do not only compare transcription accuracy; minutes quality is what actually saves time.
Fourth, pricing units differ, so don't compare raw numbers. Whisper is per-minute (about $0.006/min), Otter and Notta are monthly per-user, Tingwu and Miaojí are bundled or monthly. Per-minute and monthly can't be compared directly; convert by your volume: at 10 hours a month, Whisper API is about $3.60, Otter Pro $16.99, Tingwu about ¥99. Per-minute is cheaper at low volume, subscriptions win at scale. Estimate your monthly usage first, then convert to one unit before comparing, or a figure like "starts at $0.006" will mislead you.
FAQ
Q: Which one should a complete beginner pick? A: For Chinese scenes, Tongyi Tingwu or Feishu Miaojí (Miaojí for Feishu users, Tingwu otherwise); for English scenes, Otter.ai. All three have a UI, minutes, and a free tier, ready to use. Whisper is for developers; beginners should skip it.
Q: Which one is completely free? A: For English meetings, Otter free at 300 minutes/month (30 minutes per session); for Chinese meetings, Tingwu's free hours or Miaojí's Feishu quota; for cross-language, Notta free at 120 minutes; for unlimited free, run Whisper locally (open source, but you need a GPU and tech skills). The Whisper API has no free tier.
Q: Which is most accurate on Chinese meetings? A: Tongyi Tingwu and Feishu Miaojí tie for strongest, both natively trained on Chinese. Whisper is upper-mid on Chinese but does not do minutes. Otter is weak on Chinese; do not use it for Chinese meetings. Notta is upper-mid, workable for cross-language but not the strongest.
Q: Which one keeps data in-country? A: Run Whisper locally and data never leaves. If local deployment is too hard, pick the domestic Tongyi Tingwu or Feishu Miaojí; data stays in China. Otter, Notta, and the Whisper API all send data to overseas servers; skip them for privacy-sensitive scenes.
Q: Which one for API integration as a developer? A: The OpenAI Whisper API — $0.006/min, full docs, multilingual, can be wired to an LLM for minutes. For local deployment, run the Whisper open weights and fine-tune domain vocab. The other four are closed SaaS with less API flexibility than Whisper.
References
- OpenAI Whisper model and API pricing: https://openai.com/api/pricing/
- Whisper open-source repository: https://github.com/openai/whisper
- Otter.ai pricing and features: https://otter.ai/pricing
- Tongyi Tingwu official site: https://tingwu.aliyun.com/
- Tongyi Tingwu pricing and features: https://tingwu.aliyun.com/buy
- Feishu Miaojí official introduction: https://www.feishu.cn/product/minutes
- Feishu paid plans pricing: https://www.feishu.cn/pricing
- Notta pricing and features: https://www.notta.ai/en/pricing
- Notta multilingual support: https://www.notta.ai/en
- AI transcription tools review (third-party reference): https://www.tomsguide.com/best-ai-transcription-services