The Righteous Stress Test: Neural Frames — Dry Run, No Edits
Share
DRY RUN — NO EDITS.
Want to see how the finished music video evolved? Start with the directed Made to Spark version, then return here to compare the original unedited experiment.
For the platform features, plans and the full series of tests, return to the main Neural Frames review.
Watch both original dry runs below. Each video opens in Google Drive and is available to anyone with the link. Both are unchanged first outputs created with the app's built-in templates, backgrounds and AI characters.
Disclosure: This test is part of a paid collaboration with Neural Frames. The partner provided testing access and a designated tracking link for this fixed-fee feature; this is not our original affiliate link. Our observations and conclusions remain our own.
Before you start fixing an unfamiliar creative tool's output, learn what it produces on its own. That's the purpose of the Righteous Stress Test: complete a project using built-in templates, preserve the original, examine the result, and only then decide what deserves editing.
My first-run verdict: The built-in templates impressed me: they produced an atmospheric studio world for the full song and a distinctive performance setting for the short. I recommend exploring Neural Frames. Both original exports remain unchanged for an honest before-and-after comparison.
Try Neural Frames now for free
Free access is limited; video generation may require paid credits. Check current pricing, credits and subscription terms before purchasing.
The test setup
We used the Jack Righteous recording Made to Spark and Neural Frames' built-in background templates and AI characters. We generated one full-song landscape video and one Lyric Pro short-form vertical video. No post-generation editing or visual corrections were performed. These are baseline experiments, not polished final productions.
Generation-time context: The first render felt reasonably quick compared with other apps I have tried, but I did not time it and cannot claim a measurable speed advantage. Seeing the result helped me decide which edits mattered more than small differences in waiting time.
Test 01 — Full-length music video (3:33)
DRY RUN — NO EDITS
Landscape; 3:32.8; 1344 × 768; 24 fps. The original uses an amber-lit recording-studio world and a recurring lead character, Kwame. Its strong opening environment, equipment imagery and scene variety came from the template-based first pass. The main opportunities were creative, not a wholesale failure: introduce the song sooner, give a few recurring images clearer narrative purpose, and decide which passages need singing rather than natural musical movement. See the specific visual checkpoints below.

Watch the original full-length video (3:33)
Original unedited MP4 · Opens in Google Drive in a new tab
Test 02 — Lyric Pro short-form (0:30)
DRY RUN — NO EDITS
Vertical; 30 seconds; 720 × 1280; 25 fps. The original short establishes a recognizable performer in a distinctive concrete industrial hall surrounded by brass instruments. Strong close-ups and changing viewpoints make it visually engaging. The principal improvement opportunity is to get the performer larger on screen sooner and review extreme close-up crops; lyric and mouth-timing accuracy require a separate listening comparison to the soundtrack.

Watch the original Lyric Pro short (30 seconds)
Original unedited MP4 · Opens in Google Drive in a new tab
Try Neural Frames now for free
Free access is limited; video generation may require paid credits. Check current pricing, credits and subscription terms before purchasing.
Our seven-step method: how to repeat the test
The goal is to understand the tool before deciding what to edit. You can try this with one song and one short sample if credits are limited.
- Define: Choose a completed song and write one objective, such as “generate a vertical clip using an app-provided performer.”
- Explore: Record the workflow name, available templates, selected background and AI-character options.
- Generate: Complete the first output without post-generation changes. Check the credit estimate before starting.
- Preserve: Download and label the file DRY RUN — NO EDITS; do not overwrite it with revised versions.
- Observe: Watch the complete result once for overall feel and again for precise issues. Note timestamps.
- Identify: Separate defects (e.g., incorrect timing or changing facial features) from preferences (e.g., wanting a darker environment).
- Refine later: Prioritize a few specific changes and compare the next output with the untouched original.
Our test log: Made to Spark
| Field | Full-length video | Lyric Pro short |
|---|---|---|
| Song | Made to Spark | Made to Spark |
| Objective | Explore full-song visual generation | Explore 30-second vertical lyric-focused workflow |
| Duration | 3:32.8 | 0:30 |
| Output | 1344 × 768, landscape, 24 fps | 720 × 1280, vertical, 25 fps |
| Assets | Built-in backgrounds and AI characters | Built-in backgrounds and AI characters |
| Post-generation edits | None | None |
| Workflow / visual setup | Autopilot; studio template; Kwame, Velvet Void and Master Drive story elements visible in project editor; 11 scenes, 42 clips | Original short-form template, described as Lyric Pro in initial test; recurring performer and industrial/brass-instrument setting |
| Exact underlying AI model / version | Not documented for the original render | Not documented for the original render |
| Credits consumed | Not recorded | Not recorded |
| Generation time | Not recorded | Not recorded |
Production-data verification: The file metadata confirms durations, dimensions, frame rates and export codecs, but does not identify the generation model, exact settings, credits deducted or elapsed render time. Those figures should not be inferred from file size, video length, an account-balance screenshot or later upload timestamps. Neural Frames' Videos page documentation indicates that project records can show the model used and last updated time; its credit guide explains that costs depend on the selected Autopilot model and any rerenders. If the original project history is accessible, the original model and credits can be added without guessing.
Rerender-credit observations from subsequent editing (not original-generation totals): When I later refined the full-length studio video, I recall receiving several initial rerenders without an additional credit charge; subsequent studio clip rerenders appeared to cost approximately 40–65 credits each. In the separate short-form editor, rerenders appeared to be roughly 150–250 credits each, sometimes a little outside that range. These are approximate figures from my own project, not verified account totals, published fixed prices, or a claim that the same rates apply to every model or subscription. The original generation costs and exact render durations remain unrecorded.
Separate, verified creator estimate: In the subsequent directed full-length Made to Spark revision, the creator estimated about two hours end to end, including rendering, research, testing, note-taking and editorial work. That is a combined refinement-session estimate, not an original-render time for either dry-run export. See the refined-video case study.
Try it yourself: a ten-minute review
After generating one clip, identify one strength, one technical concern and one creative preference. Record the timestamp of each. Then decide which single correction you would test first. A useful result is a prioritized edit plan, not a perfectly generated first take.
What the original videos actually show: strengths and improvement opportunities
How to read this assessment: The approximate timestamps below were rechecked against the untouched original MP4 exports, using time-sampled frames across both video timelines. These are grounded visual and storytelling observations, not independently certified phoneme-accurate lip-sync measurements. Lead-versus-choir accuracy and exact beat alignment should be assessed alongside the soundtrack rather than guessed from stills.
Test 01 — Original full-length Autopilot video (3:32.8)
- 0:00–0:09 · Strong atmosphere, missing introduction. The deep mahogany corridor, desk lamp and amber reflections establish a striking cinematic world. As a music-video opening, however, the original does not identify the song or Neural Frames. The later directed version solved this with a title and production credit while keeping the corridor itself.
- About 0:09–0:18 · A recognizable central character. Kwame appears with the Master Drive and the same clothing, hairstyle and overall visual identity seen elsewhere. This is a strong early character anchor. Whether any particular mouth movement matches the baritone vocal requires an audio-timed check.
- About 0:24–0:54 · A convincingly unified studio setting. Console shots, the chandelier-lit room, a seated performance and handheld equipment make the environment feel intentional rather than randomly assembled. The creative opportunity is to move between those shots with greater purpose and avoid letting similar studio angles carry too much of the narrative.
- About 1:00–1:18 · Useful visual variety. The red-lit silhouettes, performer close-up and wider studio compositions introduce a change in scale and energy without abandoning the aesthetic. The first pass gives a compelling foundation for a story about creating and completing a song.
- About 1:30–2:12 · Strong world-building with slower visual passages. Backstage equipment, coiled cables, performer reactions and nearly empty spaces help communicate the work behind the music. Some of those wide or unoccupied shots slow the visual pace; whether they help at a given moment depends on the actual song section.
- About 2:18–2:48 · Cohesive but somewhat repetitive imagery. Reflective floors, red accent lighting and recurring workstation imagery keep the video consistent. The trade-off is that repeated empty-room or console compositions do not always advance the creator's story on their own.
- About 3:00–3:32 · Promising finale material. The performer, mixing console, a large speaker and finally a formally dressed group appear near the end. Those are effective storytelling ingredients, but the untouched cut leaves more of their meaning implicit. The later revision connected the group to an explicit message about creative agency.
Full-length conclusion: The initial Autopilot result gave the project an excellent, visually consistent foundation from the selected template. The best improvements were selective direction—clearer opening identity, a few message-led frames, and clip-mode corrections—not replacing the whole video. See the refined 3:33 case study for the before-and-after result.
Test 02 — Original 30-second vertical Lyric Pro experiment
- 0:00–0:03 · Excellent visual concept; small subject. The concrete atrium, staircase and clustered brass instruments make the short instantly distinctive. The opening wide view places the singer relatively far from the phone viewer, so a tighter early shot could create a faster hook.
- About 0:03–0:09 · Strong performer continuity. Side views, a frontal close-up and a walking shot retain the same recognizable clothing, character and environment. The close-up around 0:06 crops the performer's face tightly; that is a framing preference worth judging on a small screen rather than a confirmed technical defect.
- About 0:10–0:17 · Clearer music-performance emphasis. Medium and closer frontal shots give the singer more visual presence, which helps a vertical clip feel direct and personal.
- About 0:18–0:25 · Dynamic changes in scale. A striking spotlight wide shot is followed by expressive views near the brass instruments. The contrast works visually, although another distant shot can briefly reduce the character's screen presence.
- About 0:25–0:30 · Expressive closing gesture. The view returns to a wider environment before the performer opens his arms toward the end. It provides a strong visual finish that could support an actual release hook or final lyric if the chosen audio excerpt matches it.
Short-form conclusion: The unedited template provided a compelling location, consistent virtual performer and good shot variety within 30 seconds. The practical edit priorities are opening framing and careful lyric/caption review, not a wholesale visual redesign. A separate Lip Sync Pro short-form guide documents the later editor workflow, including selective segment rerenders and caption corrections.
Important audio limitation: We can confirm the visuals and exported audio track, but the above timestamps do not establish word-by-word vocal synchronization. The lead baritone and choir overlap in the song. The creator's later hands-on review identified passages where the character needed to follow the lead only, or stop mouthing lyrics and simply nod to the music. Those edits are described in the revised-video articles, not attributed to changes in these untouched baselines.
An unexpected advantage of testing a visual performance: seeing an avatar interpret the recorded vocal also drew my attention to pronunciations and phrasing in the original Suno song. The process helped me evaluate not only the video but how the lead vocal might need to be presented, particularly when the baritone and choir interact. That observation informed the later edits; it should not be mistaken for a flaw proven by still images.
What we changed after this baseline assessment
Keeping the two originals unchanged made the comparison useful. The later full-length project reused the existing cinematic setting and selectively added the opening title, creator-focused visual statements, and more appropriate performer motion. For instrumental passages, switching a clip from Lip Sync to Cut allowed a normal head-nod motion instead of trying to animate nonexistent vocals. The short-form follow-up concentrated on segment-by-segment rerendering and caption corrections. These decisions were straightforward to make in the editor and preserve the initial test as an honest reference point.
Takeaway
Keep the original result, compare what improves, and use that evidence to direct the next edit.
Read our Neural Frames app overview.
Create What You Love | Love What You Create.