What it does
Run your lines against a scene partner that catches every flub, so you walk into the audition, class, or rehearsal knowing it cold and trusting that you do.
Currently
Live MVP. Browser-only, no signup, your audio never leaves your device. Load any script, pick your character, and the computer plays everyone else, listening as you run lines, flagging mistakes, and rewinding to the trouble spots. Privacy-safe instrumentation is the next thing in.
Roadmap
/now-next-later-roadmapNow/Next/Later Roadmap- Privacy-safe instrumentation baseline
- Instant-start (library scene to rehearsing in one tap)
- Blended voice plus tap confirmation (recognition robustness)
- Mic-permission and tap-mode fallback UX
- Guided first run
- AI-voice key strategy (gateway, fallback, free tier)
- Local saved scripts and resume
- Memorization ladder (the v2 flagship feature)
- Readiness / confidence score
- Clean-run streak and badge
- Focused drill mode (loop the weak chunk)
- Verifiable privacy badge
- PWA polish, installable, offline
- Shareable scene links
- Library / SEO expansion drumbeat
- Accessibility-first rehearsal
- Spaced repetition across sessions
- Opt-in privacy-safe reminders
- Flub heatmap
- Classroom / teacher kit
- Musical theatre song mode
- Self-tape recording (deferred, off-wedge)
- Native iOS/Android app (deferred, PWA first)
Recently shipped
- Jun 24, 2026Added the instant-start 'Try it now' hero so a first-timer can load a library scene and be rehearsing in one tap, no setup, no mic decision up front.
- Jun 10, 2026Shipped the MVP: load a script, pick your character, run lines hands-free while the tool detects paraphrases, skips, insertions, and blanks at the word level and rewinds you to the spot.
How it's built
How I built it
The flub is the product
Every line-running app I looked at does the same thing: a robot reads the other character's lines so you can practice the back-and-forth alone. That is table stakes, and it is genuinely useful, but it is not where the pain lives. The hard part of memorizing is not the first read-through. It is the third and fourth, when you half-know a scene but keep paraphrasing the same line, skipping the same beat, blanking in the same spot, and re-reading the whole thing hoping it sticks. So I built Line Runner to listen. It plays everyone else, and as you run your lines it catches the exact words you flubbed at the word level, paraphrase, skip, insertion, or blank, and rewinds you straight to the trouble spot. The mistake-drill is the wedge. The cue reading is just the price of entry.
What I cut, and why
The strategy run scored the whole feature pool, and the math was blunt about it. Self-tape recording and a native app, the two features every competitor markets hardest, sank to the bottom on effort and landed firmly in Later. That felt right rather than reckless: self-tape is off-wedge and privacy-complicating, and a native app is a twelve-week build when a polished installable web app closes most of the mobile gap. What rose to the top instead was unglamorous foundation work, an instant-start path so a first-timer is rehearsing in one tap, and making recognition robust enough that a missed word never blocks a run. The headline new feature, a memorization ladder that walks you from prompted to off-book, is sequenced right behind that foundation so it launches onto a base that actually activates and that I can measure.
The bet
Line Runner is free, no signup, and your audio never leaves your browser, and that is a strategy, not a giveaway. The cheapest growth for a tool like this is a conservatory cohort that shares everything and a working actor who tells the rest of the cast, and both of those die the instant you put up a paywall or an account wall. So the North Star is not visitors or signups, it is Weekly Clean Runs: scenes you got through cleanly without prompting, the measurable version of "I know this cold." The pre-mortem named the real way this fails, which is bursty usage with nothing to pull you back between projects, and so the work I obsess over is retention inside a project window: local saves, a readiness signal, a clean-run streak. Catch the flub, drill the spot, trust that you know it. That is the whole product.
The strategy behind it
A snapshot of the product process this app was built from. Each artifact comes from a k8 skill you can run yourself. It reflects the original plan, not the live state above.
Jobs to be done
/jtbd-analysisJTBD Analysis| When I… I want to… | Importance | Frequency |
|---|---|---|
| When I have an audition, class, or off-book rehearsal date coming, I want to reliably memorize my lines word-for-word in the time I have, so I can walk in knowing it cold. | high | high |
| When I need to run a scene but no scene partner is available, I want the other characters' lines delivered to me on cue, so I can practice the back-and-forth alone, anytime. | high | high |
| When I half-know a scene but keep paraphrasing, skipping, or blanking in the same spots, I want to be told precisely where I went wrong and rewind straight to it, so I can drill the weak spot instead of re-reading what I already know. | high | high |
| When I am about to perform or audition, I want proof I can get through it without prompting, so I can stop white-knuckling and actually act. | high | high |
| When I am still rough and embarrassed about it, I want to rehearse alone without burning a favor or looking unprepared, so I can fail safely until I am good enough to run it with a human. | high | medium |
| When I rehearse, I want to be held to the playwright's actual words and not let myself drift, so I can honor the text and not get flagged by a director, writer, or stage manager. | high | high |
Persona
/persona-createPersona Create- Be off-book and confident for the audition
- Never fumble on camera
- Reuse one tool across a constant stream of new material
- No scene partner available on demand, often rehearsing alone at 11pm
- Paid cue-reader apps gate the useful parts behind a subscription
- Robotic text-to-speech pulls him out of the scene
- Paraphrases under pressure and does not notice until a director flags it
Persona
/persona-createPersona Create- Show up to scene study off-book
- Never be the one holding the class up
- Build the lifelong habit of word-perfect work her training demands
- Cannot afford subscriptions
- Roommates as scene partners are unreliable
- Wants to drill the chunk she keeps blanking on, not re-read the whole scene
Persona
/persona-createPersona Create- Hit the director-set off-book deadline without dread
- Rehearse in the small windows her busy life allows
- Not let the cast down
- Memorization gets harder with age and a busy life
- No rehearsal partner at home
- Existing apps feel built for pros and ask her to sign up and pay for a hobby
North Star metric
/north-star-metricNorth Star MetricThe number of scene runs (or substantial chunks) completed end-to-end at or above the accuracy threshold, summed across all users, per week. A clean run is the measurable proxy for 'I know this cold.'
RICE prioritization
/rice-scoreRICE Score| Feature | Reach | Impact | Conf. | Effort | Score |
|---|---|---|---|---|---|
| Instant-start (library scene to rehearsing in one tap) | 10 | 2 | 80% | 1 | 16 |
| Blended voice plus tap confirmation (recognition robustness) | 9 | 3 | 80% | 2 | 11 |
| Mic-permission and tap-mode fallback UX | 9 | 1 | 80% | 1 | 7 |
| Memorization ladder | 8 | 3 | 80% | 3 | 6 |
| Local saved scripts and resume | 8 | 2 | 80% | 2 | 6 |
| Self-tape recording mode | 5 | 2 | 50% | 8 | 1 |
| Native iOS/Android app | 7 | 2 | 50% | 12 | 1 |
Pre-mortem risks
/pre-mortemPre-mortemBursty usage (no lines to learn between projects) plus no account-based hook means retention flatlines and Weekly Clean Runs decay to near-zero between project seasons.
highMitigation: Local saved scripts and resume, opt-in privacy-safe reminders, and a per-project-window success view. Accept some churn as structural and optimize per-window delight.
Speech recognition fails on non-American accents, noisy rooms, and verse, so the aha moment becomes the abandonment moment and the differentiator silently excludes a chunk of the market.
highMitigation: Blended voice plus tap confirmation in the Now wave so a recognition failure never blocks a run, plus testing across diverse voices and rooms.
Nobody finds it: a web tool with no App Store, slow-to-compound SEO, and a single maker juggling consulting and a job search means the roadmap stalls at the foundation wave.
mediumMitigation: Treat the library/SEO drumbeat, share links, and honest social proof as the growth model itself, not optional polish. Under-investing here is the risk.
The privacy claim is technically overstated: saying 'your audio never leaves your device' while the browser's recognizer is cloud-backed could become a trust scandal.
mediumMitigation: Precise copy that says exactly what is and is not sent, plus a data-flow audit so the words match reality. A trust breach here is existential for the brand.
Mistake-drill demand is not actually real, and the wedge the whole strategy orbits turns out to be unvalidated.
mediumMitigation: Validate via usage instrumentation (drill-mode engagement, rewind usage) rather than over-building the drill branch up front. The memorization ladder hedges by serving the proven get-off-book job.