Text to speech with no character limit.

On a paid plan a single narration job accepts up to 1,000,000 characters — a whole book in one paste. Anything past 20,000 characters is handed off as a background job rather than synthesized in the request, so nothing times out and nothing is silently truncated: the text is split into chunks, narrated, and stitched back into one continuous MP3. Free accounts are capped at 20,000 characters per job.

Roughly 25,000 characters becomes about 30 minutes of narration, so a 300,000-character book lands around six hours and a million characters is close to twenty. It arrives as one file you download and keep.

No, and the distinction matters. Length is unmetered; characters are metered. Narration costs 1 credit per 10 characters, doubled on the six premium HD voices, drawn from the same monthly balance that pays for images and paid video clips. So Basic's 5,000 credits is about 50,000 characters a month if you spend the whole balance on audio, and Studio's 200,000 credits is about two million. What you do not hit is a per-job ceiling.

Yes. The same voice, style, pitch and pace are applied to every chunk, and the chunks are stitched rather than re-synthesized, so there is no mid-book voice change and no seam at the joins.

51 voices in the main picker plus 34 more in the full library (85 in all), across 16 languages. Twelve emotional styles are available on 18 expressive voices — no single voice supports all twelve, and the richest supports ten plus default — and pitch and pace run from −50% to +50% on every Azure voice.

Two, explicitly. Hebrew and Thai have no voice we will route to, so that text is refused before any provider call with a message saying so, rather than being mis-narrated in the wrong voice. Russian works, but only if you pick a Russian voice — we do not auto-switch a script into a language nobody asked for. Non-Latin scripts route to the dedicated voice for that script. The 13 OpenAI voices are the exception: they are multilingual, so they are never re-routed or refused — they read the script in the voice you picked.

Yes. POST /api/v1/tts accepts up to 1,000,000 characters: short texts return audio directly and long texts return a job you poll for the finished MP3. Billing is identical — 1 credit per 10 characters, minimum one credit per request.

That is what the rest of Vidsly is for. The same script that narrates here can be sent through the video pipeline: a scene image per beat, motion generated from each image, a music bed, and a finished MP4 in 9:16 or 16:9. On the unlimited lane that whole generation costs zero credits per clip, from $29 a month.