How it works
No dashboards, no setup. Here's the exact pipeline, and below it a full real sample report so you can see precisely what you get before you pay.
A PNG or JPG screenshot, or a short MP4 clip if you want audio scored too. Up to 10 MB image, 50 MB / 30 sec clip. We extract a still in memory and never write your bytes to storage.
One LLM pass reads the frame against a fixed rubric and returns lighting, framing, audio, and scene-composition scores, each 0-100, each with one affirming observation, the single finding that matters, and a named fix.
From all the findings it picks the three changes with the best impact-per-dollar-per-minute, ranks them, and tags each with cost and time so you know what to do first.
Renders instantly, lands in your inbox, and stays available to revisit. Every report is labeled AI-generated and not human-reviewed.
A full real sample report
The report below is genuine Streamlint output for a Just Chatting setup lit by a side window, scored 71/100 overall. Same format, same voice, same depth your own audit returns.
WORKINGThe side window gives you soft, directional light with a natural catch-light in the eyes, which already looks more flattering than most ring-light-only setups.
FINDINGBecause the window is your only source, the side of your face away from it falls into a noticeable shadow, and the color shifts cooler whenever a cloud passes.
FIXAdd a $30 Elgato-style key light or a cheap LED panel on the shadow side, set just below window brightness at 5000K, sitting on a shelf to your side facing you. That fills the dark cheek and steadies the color when the daylight changes.
WORKINGYou're framed at chest-up with good headroom and the rule-of-thirds placement is close to ideal, so the shot already feels composed.
FINDINGYou sit dead-center, which is fine for a talk stream but leaves the chat box and your face competing for the same middle column.
FIXShift your chair about six inches to your left (your left, seated) so you sit on the left third. That opens clean room on the right for chat or a topic card without crowding your face.
WORKINGFor a Just Chatting stream you've clearly thought about the visual read, which usually means you care about the sound too, you just haven't sent it yet.
FINDINGAudio scoring needs a 15-30 second clip with sound. We couldn't grade audio from a still frame alone.
FIXReply to your report email with a 15-30 second talking clip and we'll grade levels, clipping risk, and room echo, then name the one mic or treatment fix that moves the needle. Included in your $29.
WORKINGRestraint is the strength here: a minimal overlay and the plants behind you give the scene a calm, lived-in identity instead of neon clutter.
FINDINGThe webcam border and your name plate use a low-contrast gray on a bright background, so they nearly disappear and the scene reads slightly unfinished.
FIXBump the name-plate text and webcam border to a single brand color with a subtle drop shadow in OBS. It costs nothing and makes the minimal look intentional rather than empty.
The three quick wins
Shift your chair to your left onto the left third and open the right side for chat (see Framing fix). Free and instantly more professional.
Add a $30 fill light on the window-shadow side (see Lighting fix). Evens your face and stops the cloud-driven color shift.
Recolor the name plate and webcam border to one brand color (see Scene fix). Makes the minimal overlay look deliberate.
AI-generated, not human-reviewed — here's what we caught.
Questions
Lighting, framing, and scene composition are highly readable from a single still, that's what a viewer sees in their first two seconds. Audio is the exception: it needs sound, so a screenshot scores audio as pending and invites a short clip for a free audio pass. Clips score all four dimensions in one shot.
Whatever fits the frame and the budget. The rubric biases to fixes under $30 and under 30 minutes first, names the exact product and setting (for example a $22 ring light at 5500K), and only ranks a bigger upgrade when it unlocks a 30-plus point jump.
No. When a fix touches something personal, fan gear, posters, a VTuber model, it reframes rather than removes. Your identity is why viewers pick you, so the tool works around it.
About 60 seconds from upload to a report that renders on the page and lands in your inbox.