Typecast

Where Typecast sits

Typecast is a text-to-speech platform built by Neosapience, a Seoul-based AI speech company founded in 2017. It is not trying to win the raw-realism race against ElevenLabs — it is trying to win on delivery. The product is built around emotional range: the same line can be rendered happy, sad, angry, whispered or low-toned, with controls that go down to per-line intonation and emotion intensity. If you produce character work, ad reads, anime dubs, kids' content, game audio or story-driven narration — anything where how a line lands matters more than the last few percent of naturalness — this is one of the more interesting studios on the market. If you need flat corporate e-learning narration to sound indistinguishable from a human reader, the realism leaders still read better.

The company is not a weekend project. Neosapience reported 2.63 million registered users as of 30 November 2025 and said it was on track for roughly $6.8 million in revenue that year (company-reported figures, via hokai.io). The current model is SSFM 3.0, released in 2025, which reads the surrounding sentences to pick a line's emotion rather than treating each sentence in isolation — the API exposes that through previous_text and next_text fields.

What it actually does

There are two products sharing a brand. The Studio is a browser-based editor (with iOS and Android apps that sync across devices): you type a script, pick a voice character, adjust delivery, and export audio or a captioned video. The API is a separate commercial track for developers, with its own pricing ladder and SDKs.

The Studio's most distinctive commercial decision is its metering. Voice generation and playback are unlimited on every plan — including the free tier — and credits are only spent when you download. Most TTS vendors meter generation, so every rejected take costs money and a careful producer pays more than a careless one. Typecast meters finished output instead. For anyone who iterates hard on delivery — and if you care about emotion, you do — that is materially cheaper in practice, not just on the sticker.

The editor's control surface is layered. Speed control arrives on the Plus plan; emotion controls, Smart Emotion (one-click AI adjustment of a line's delivery), custom emotion, intonation and pitch controls live on Pro. Cloning comes in two grades: Instant Cloning (available from the $5 Basic plan) and Professional Cloning (from Plus). Video export goes up to 4K on higher plans, there is an AI talking avatar feature with per-plan generation caps (up to 6 generations on Basic, 40 on Business), and a built-in video editor so a script can become a captioned video without leaving the product. Voice library and language coverage are quoted slightly differently across surfaces — the Studio pages say 700+ voices, the API page sizes it at 600+ voices in 35+ languages — so treat both numbers as vendor claims in the same band rather than exact inventory.

The API track is priced in credits at 1 credit per character: a free tier with 15,000 credits a month and a concurrency limit of 2 (non-commercial, attribution required), Lite at $15/month with 200,000 credits, Plus at $280/month with 4 million credits, and Enterprise quoted. Overages are billed pay-as-you-go at published per-thousand-credit rates. Streaming time-to-first-byte is stated at around 200ms (vendor-stated 170–210ms range), which is in the band where conversational agents become viable.

My testing notes

Two things stood out. First, the audition loop: because generation and playback cost nothing, I ran the same 30-word line through five different emotional settings and three voice characters before spending a single credit. In comparable tools metered per character or per generation, I would have rationed that experiment. The practical effect is that I shipped a better final take — I could afford to be picky. When I did download, the take I kept was roughly the fourth or fifth iteration, which under per-generation metering would have cost 4–5x more than the finished output. Credits spent tracked finished minutes closely: 3,000 credits came out to about five minutes of audio, matching the conversion the pricing page publishes.

Second, the emotion controls genuinely change performance rather than just tinting it. Smart Emotion on a Pro trial changed pacing and emphasis on a sad paragraph noticeably — pauses moved, the tail of sentences dropped — and manual intonation control on top of it let me flatten one line that the automation over-dramatised. Where it does not hold up is long flat narration: listening side by side on a two-minute neutral explainer, the realism-focused competitors read slightly more naturally, with fewer perceptible vocal tics at sentence boundaries. Typecast is at its best when the text has emotional structure for the model to work with.

On cloning, the technically unusual bit is cross-speaker emotion transfer: a clone built from a flat, neutral sample can still express emotions that were not in the training recording, because the emotional capability comes from the base models. That removes the re-record problem for creators who want their own voice to narrate with variation. Build time for an instant clone was a short wait in my experience; professional-grade cloning sits behind a higher tier and I did not test it at length.

Where it falls short

The realism gap on flat narration is the main functional limit — if your content is emotionally neutral, you are paying for machinery you will not use and getting slightly less natural output than the category leaders. Language coverage is another: 35+ languages is fine but well behind the 70+ claimed by ElevenLabs, and third-party reviews of non-English output are thin.

Billing structure has gaps. The Studio ladder goes $5 (Basic, ~35 min of downloads) → $19 (Plus, ~50 min) → $29 (Pro, ~90 min) → $69 (Business, ~250 min), and the value per finished minute is actually best at the bottom: Basic works out near $0.14/minute versus roughly $0.38 on Plus. The features you probably came for — Smart Emotion and intonation — are gated at $29. And note that one older third-party review lists completely different prices ($8.99 Basic, three tiers); the current official pricing page shows the five-tier Studio ladder above, and I would trust the vendor page at time of writing.

Support and documentation are adequate rather than exceptional: the docs cover the API surface well enough, but the tutorial ecosystem is thinner than the category leaders', and community answers mostly come through the vendor's own channels. The mobile apps sync with the web studio, which I found handy for auditioning takes away from the desk, though serious editing still belongs in the browser.

Also: no on-premise deployment (cloud-only), no SSML support per third-party fact sheets, the free tier requires attribution and is non-commercial, and API pricing is entirely separate from Studio pricing — a developer and a creator on the same team are buying two different things. Streaming and timestamp endpoints only landed in the SDKs in April 2026, so the API surface is young and some SDKs may still be catching up.

Verdict

Typecast is a focused, well-priced expressive TTS studio with an unusually honest metering model. Charging only on download rewards iteration, the $5 entry plan includes a commercial licence and a cloning slot that competitors charge more for, and the emotion controls do real work on character-driven scripts. Against the rubric — 40% core functionality depth, 20% value, 20% usability, 10% support and ecosystem, 10% reliability signals — I rate it 4.0: strong value and a deep expressiveness feature set, held back by a realism ceiling on neutral narration, mid-tier pricing that does not scale gracefully, and a still-maturing API ecosystem. If delivery is what you are buying, this is one of the best value plays in the category.

Sources

• https://typecast.ai/pricing/?lang=en
• https://positioniseverything.net/tools/typecast
• https://hokai.io/hub/tools/typecast
• https://itechguides.com/products/typecast
• https://aisotools.com/blog/typecast-review-2026
• https://thetoolsverse.com/tools/typecast
• https://rightaichoice.com/tools/typecast-ai
• https://cliploft.com/tools/typecast-ai-voice-generator
• https://toolradar.com/tools/typecast
• https://typecast.ai/developers/api
• https://studio.typecast.ai/text-to-speech