Every FrameForge calculator, on one page

How to read: 70+ means your hook is structurally tight. 50–69 is fine but improvable — usually one of the cheap fixes in the action list moves you up a tier. Under 50, you're competing on thumbnail CTR alone, which is an uphill battle.

Limits: This is a heuristic for YouTube-style vertical-feed retention. TikTok and Shorts favor stronger visual motion in the first 0.5s — bump the motion slider up by 2 if that's your target. The score does not know your niche, your thumbnail, or your audience size — none of which can be solved inside a calculator.

How to read: The percentages on the X-axis are where each named beat should land. The Y-axis is "emotional intensity" — a rough guide, not a script. If your climax lands before the 70% mark, the back third of your video is going to feel like a victory lap.

Limits: Templates are starting points. Comedic and educational videos often want a faster midpoint (around 35%) and a delayed climax. Treat the table as a checklist, not a ruler.

How to read: "Shots to capture" is your minimum viable coverage — it assumes you'll use ~70% of what you shoot (the rest is coverage insurance, mistakes, or unused). "Per subject" helps you decide whether to leave for a new location or shoot more at the one you have.

Limits: Doesn't account for animation-heavy channels (where B-roll is replaced by motion graphics — fewer shots, longer per shot) or pure talking-head podcasts. For those, set the talking-head ratio to 95%.

How to read: Score blends the five levers. CTR estimate is a baseline — small channels will sit below it, large channels above, because thumbnail is a multiplier you can't see here.

Limits: No title tool can predict whether your audience will trust you. "Clickbait" titles score high here on paper; the long-term cost shows up in retention and session-watch, not the title itself.

How to read: YouTube requires ≥3 chapters and the first one ≥10 seconds. This tool enforces both. It also flags duplicate titles and gaps >10% of runtime.

Limits: Chapter names are a UX call, not a math call. Keep them <50 chars so they don't truncate on mobile.

How to read: Each card shows the same scene cropped to that ratio. If your subject disappears in 9:16, that's a composition problem you can fix by re-framing in-camera.

Limits: Doesn't account for digital stabilization crop (typically 10–15% extra). If you're shooting handheld, leave 15% padding on each edge.

How to read: Yellow = caption area (keep critical visuals above it). Red = UI overlap (likes, follow button, share). Green = safe center (where eyes should land).

Limits: UI layouts change. TikTok has moved its caption strip 3 times in 24 months. Always re-check the platform you're publishing to before locking the framing.

How to read: "Look" lets you push the white balance off-neutral for creative effect — push it warmer for cozy vibes, cooler for sterile/digital.

Limits: Mixed lighting (window + lamp) breaks this tool. Solve mixed light at the source: turn off what doesn't belong.

How to read: Angles are from the camera-subject axis. Distances are in head-widths (a useful on-set ruler that scales with your subject).

Limits: Real rooms have ceilings, walls, and windows that constrain placement. Use the diagram as a starting point; if you can't hit the angle, soften the light source to compensate.

How to read: Distance and gain are coupled — moving the mic further forces more gain, which also amplifies noise. The "Min distance" is the closest you can get while still leaving headroom.

Limits: Doesn't model reflections (a hard-walled bathroom behaves nothing like the math). For untreated rooms, halve the recommended distance.

How to read: Kelvin handles the global warm/cool axis. Tint handles the magenta/green axis (often more powerful than people realize). The Lift/Gamma/Gain numbers are DaVinci Resolve style offsets — paste them into your color page.

Limits: This is a starting grade. Real cinematic color has 3-5 nodes, plus a parallel node for skin tone protection. The palette is for mood-board reference, not hex-perfect grading.

How to read: Going from 60 → 24 at 100% speed is a 2.5x slowdown (cinematic). The 180° shutter equivalent helps you preview the motion blur amount.

Limits: Optical flow / motion interpolation can do better than this naive math but introduces artifacts. For talking-head interviews, prefer optical flow. For action, prefer the naive numbers.

How to read: "Recommended" is a sweet spot. "CBR target" is what you'd put in a streaming encoder (Twitch, Vimeo). "Min" is the floor — going below produces visible macroblocking on motion.

Limits: Real-world encoders vary ±15% from these numbers. Always do a 30-second test render before committing a 2-hour export.

How to read: A LUT designed for log footage should be applied at <50% on a Rec.709 source. A LUT designed for the same gamma should go at 75–100%.

Limits: "Skin protection" means dropping a parallel node with a skin-tone qualifier that masks the LUT's effect on faces. Skipping this turns people orange.

How to read: Use the lightest tone for backgrounds, mid tones for subjects, and saturated tones for accents/CTAs. Monochrome works best for serious/educational; triadic works for energetic entertainment.

Limits: A palette isn't a grade. The colors here are reference — your actual grade depends on the source footage.

How to read: "Quality retention" is a VMAF-style estimate (0–100). Below 80, your re-encode is visibly worse than the source. "Perceived loss" is the relative score.

Limits: ProRes is treated as a lossless intermediate; actual retention depends on the source. For archival, use a mezzanine codec (ProRes 422 HQ or DNxHR HQX) — never the delivery codec.

How to read: Apply gain first, then check true peak. If pushing gain pushes you above -1 dBTP, you need a limiter, not more gain.

Limits: Loudness normalization is platform-applied at playback — YouTube will turn down a -10 LUFS video. For consistent loudness across platforms, target the loudest platform's spec (-14 LUFS).

How to read: The 4 bands are: high-pass (rumble), low-mid (warmth), presence (clarity), and air (sparkle). Q is the bandwidth of each cut/boost — narrower Q = surgical, wider Q = musical.

Limits: This is a starting preset. Real voice EQ depends on the actual recording — do a spectrum analysis on your audio and subtract what's already there.

How to read: "Δ from voice" is the gap you should leave between music and voice. Bigger gaps feel safer; smaller gaps feel more energetic. Sidechain amount is a compressor parameter — paste into your DAW.

Limits: Sidechain ducking only works on music with steady mid-range energy. For sparse cinematic music, manual volume automation beats any compressor.

How to read: 30 ms drift is the threshold for "I can hear it." 60 ms drift is the threshold for "anyone will notice." If you're past the perceptible threshold, apply the stretch ratio to the audio track.

Limits: Drift accumulates linearly only if the source is consistently bad. Real wireless drift is temperature-dependent — cold weather makes it worse.

How to read: CPS above the language threshold is technically readable but feels rushed. Long subtitles (over ~42 characters per line) need a 2-line layout, not a single-line one.

Limits: CPS ignores comprehension difficulty. A line like "He hypothesised that..." reads slower than "Hello world." even at the same character count.

How to read: Music videos can sustain cuts every 0.5s. Tutorials should sit at 4–8s/cut or you confuse the viewer. Vlogs are flexible.

Limits: Cut density ignores "soft" transitions (dissolves, wipes). A dissolve every 3 seconds counts as a cut here, but feels very different.

How to read: 50–60% talking head is the sweet spot for YouTube essays. Vlogs can push 75%. Pure talking-head podcasts sit at 95% — but they live on watch-time, not session-watch.

Limits: B-roll quality matters more than quantity. One well-placed 3-second cutaway every 20 seconds beats 20 random 1-second clips.

How to read: The pacing score is a weighted blend. "Biggest lever" tells you which single input, if you fixed it, would lift the score the most.

Limits: Pacing is content-aware — a 30-minute documentary has different pacing targets than a 30-second TikTok. Use the score comparatively, not absolutely.

How to read: Going 16:9 → 9:16 typically loses 75% of your horizontal pixels. Subject offset tells you how far to pan your crop to keep the subject centered.

Limits: For moving subjects, manual keyframing is still better than any auto-reframe. This tool helps you plan; the timeline still needs your hands.

How to read: Codec/container/bitrate is the minimum to satisfy the platform. For archival, also export a ProRes master.

Limits: Platforms change specs. YouTube removed the 480p option in 2024. Re-verify with the official upload docs once a year.

How to read: Use the layout overlay in your NLE editor to pre-place your end-screen elements. YouTube's editor accepts up to 2 clickable cards + 1 subscribe button.

Limits: YouTube end-screen elements must be added in YouTube Studio after upload — this tool only helps you plan the visual layout.

How to read: A high CTR comes from extreme contrast + clear subject + 1–3 words max. Faces outperform objects in nearly every niche YouTube has measured.

Limits: Doesn't measure title-thumbnail coherence. A "what I learned" thumbnail with a "scary moment" title still loses to a coherent pair, even if the thumbnail scores high here.

How to read: Questions outperform statements 5:1 in comment volume. Polls drive more replies than open-ended prompts in most niches.

Limits: "Predicted comments" assumes a baseline channel size of 10k subs. Adjust mentally: 100k channels multiply this by ~3, sub-1k by ~0.3.

How to read: The dotted line is your predicted retention. The shape matters more than the absolute numbers. A steep cliff at 30s = weak hook. A long flat plateau = strong middle. A drop at 80% = weak ending.

Limits: This is a heuristic. Real retention depends on your niche, your audience's expectations, and even time of day.

How to read: Median = what most top performers do. Sweet zone = where watch-time-per-view is highest. Min/Max = the floor and ceiling that still rank.

Limits: These are aggregate signals, not rules. A 4-minute essay can outperform a 14-minute one if the niche expects short form.

How to read: A 20-min essay on a Pro GPU with light edits should render in 4–8 minutes. Heavy VFX on CPU only can take 3+ hours.

Limits: Doesn't model disk speed (a slow SSD can bottleneck more than GPU). Always do a 1-minute test render first.