CutBay Review 2026
Higgsfield Black Friday: up to 65% off, ends 18 Nov. See offer ◆
CutBay is a chat-driven “video agent”: you give it a prompt, a web page, an image, a document or a slide deck, answer a few questions, and it builds a short video with narration, music and captions. We ran 22 jobs over five sessions on 28 and 30 Oct 2026. Web-page, prompt and image jobs came out as tidy 1080p videos without a watermark. The headline deck and document workflow didn’t hold up: every PowerPoint and document job we ran with a voiceover exported silent. It also isn’t a cinematic generator or an avatar tool.
Some links are affiliate links. We may earn a commission at no extra cost to you. Disclosure.
Disclosure: CutBay gave us an upgraded paid account (Ultra plan) for this test, and AI Video Lab has a link exchange with CutBay. CutBay didn’t review or approve this page, and neither arrangement affects our score or ranking.
Test stills (28 & 30 Oct 2026)

Prompt-to-explainer test (30 Oct): clean vector-style scenes with voiceover and music. The burned-in captions and on-screen title appeared even though we asked for no subtitles.

URL-to-video test (28 Oct) using our own homepage. CutBay scraped the page and pulled its facts through accurately, but the video ran 31.8 seconds when we asked for 20.


Stills from our own exports on the CutBay account at [email protected]. They show the output style, not a full quality test.
What We Tested
- Ran 22 jobs over five sessions on 28 and 30 Oct 2026: text prompts, a motion-graphics title, four PowerPoint decks and documents (.pptx, .md, .docx), three URLs, an image-based product ad, a template remix, narrated explainers, a vertical 9:16 video and two talking-head tests.
- Logged the credits, generation time, failures, retries and agent stalls for every job.
- Checked every export file for resolution, length, audio track, captions and watermark.
- Read the export options in the editor and both billing toggles on the public pricing page.
Pricing last checked 30 Oct 2026 (cutbay.ai/pricing, Monthly and Annually toggles). Product details tested 28 and 30 Oct 2026 on the Ultra plan. AI tools change quickly, so always confirm current limits before buying.
Overview
CutBay is built around a chat agent rather than a timeline. You describe the video or attach source material, and the agent asks follow-up questions about length, aspect ratio, voice, music and subtitles. It then writes a brief, generates the scenes and opens them in an editor where you can refine by chat, step through scenes, undo and export. The app also has a template-based Motion Graphics mode, a Talking Head Editing tool for footage you upload, and an Explore library of templates you can remix.
It’s important to know what “text to video” means here. CutBay mostly assembles stock footage, AI-generated stills and animated motion-graphics scenes. It isn’t a generative video model like Kling or the Seedance and Veo models on Higgsfield. That suits explainers and simple promos, and it doesn’t suit filmic shots.
Reliability (rounds 4 and 5, 30 Oct 2026)
Our first three sessions showed an agent that often stalled until nudged and two silent exports. In the fourth session we ran eight jobs with the same request each time (narration, music, captions and a set length) to measure how often things go wrong.
| Measure | Round 4 result |
|---|---|
| Failed runs | 4 of 8 failed at “Designing scenes”. One chat message asking for a retry fixed each one, but the failed run’s credits were not returned, so those jobs cost roughly double. |
| Silent exports | 4 of 8: all four PowerPoint and document jobs exported with no audio track and no captions, even though voiceover and subtitles were selected. |
| Agent stalls | 2 of 8 needed nudges (3 in total), one with leaked internal error text. The agent also asked whether we wanted to review the scene plan in 3 jobs after we’d asked it to skip questions. |
| Length followed | 4 of 8. The deck and document jobs hit 20.0s exactly; the URL, image and prompt jobs overshot by 20–64%. |
| Captions followed | 5 of 8 |
| Resolution and watermark | All 8 exported at 1080p with no watermark. |
| Credits | 414.5 for 8 jobs, about 51.8 per job including the charged failures. |
The pattern: in our tests, document-to-video exports (PowerPoint and Word/Markdown) came out silent, with no audio track and no captions, after a failed first attempt. Across rounds 2, 4 and 5, every PowerPoint and document job with a voiceover (8 of 8) exported silent, and every one of them had failed or stalled first and been retried. The web-page, image and prompt jobs all exported with sound. We can’t tell whether the bug sits in the retry path or in the deck and document templates.
Round 5 (one job at a time): to rule out running jobs in parallel, we re-ran one PowerPoint deck and one .docx document strictly one after the other. Both failed again at “Designing scenes”, needed one chat retry (charged again: 25.5 and 27.9 credits in total), and both exported at 20.0s and 1080p with no audio track and no captions. The document job also stalled once until nudged and showed the agent’s internal reasoning text in the chat. So running jobs at the same time doesn’t explain the failures or the silent exports.
In round 4, J1–J6 ran at the same time, so their per-job credits can’t be separated and are reported as a batch.
Test Log
Rounds 1–3 (28 and 30 Oct). Credits come from the account balance before and after each job. All exports used MP4, Standard quality, 30fps and 1080p unless noted. Times are approximate.
| Test | Credits | Generation time | Result |
|---|---|---|---|
| Prompt: red kite at sunset, 5s (28 Oct) | 12 | ~2m40s, after several minutes of questions | Painterly, near-still image with slight motion and a string artifact. Added a title we didn’t ask for. |
| Motion Graphics title (28 Oct) | ~14.3 | ~1m15s | Crisp, clean template. Ignored our background and “no music” requests. |
| Prompt: cyclist at night, 20s, stock media (28 Oct) | 23.9 | ~5 min, after ~4 rounds of questions | Real stock clips with genuine camera motion, cut into 4 scenes rather than one tracking shot. One AI-looking frame had a stray doodle artifact. Music present. |
| PowerPoint to video (28 Oct) | 26.7 | ~9 min, incl. a stall | Accurate, cleanly redesigned slides. Voiceover was charged, but the export had no audio. |
| Document (.md) to video (28 Oct) | 24.8 | ~4 min, incl. 1 failure and retry | Correct content and small captions. Export had no audio. Credits were charged for both attempts. |
| URL to video: aividlab.co (28 Oct) | ~69 | ~6.5 min | Accurate facts from the page, voiceover and music. Ran 31.8s against the 20s we asked for. |
| Template remix: business promo (28 Oct) | 20.7 | ~2m20s | 30s corporate film with a template soundtrack. Fast, and no stalls. |
| Prompt: rotating bottle ad, 10s (28 Oct) | 27.7 | ~4m50s, incl. a stall | 11s. Flat shape instead of a 3D turntable and weak studio lighting. Clean end title. |
| Narrated explainer: rainbow, 15–20s (30 Oct) | 39.3 | ~8m45s, incl. a stall | Audio track present (voiceover and music). Ran 27.8s and added captions despite “no subtitles”. |
| Chat agent asked for a lip-synced avatar (30 Oct) | 18.7 | ~5m50s, incl. a stall | No avatar. A static illustrated woman with voiceover and captions, and no lip or body movement. Ran 16.8s against ~10s asked. |
| Talking Head Editor on our own 10s clip (30 Oct) | ~0 (no visible charge) | ~40s | Correct subtitles added. The footage passed through unchanged, with no re-sync. |
| Vertical 9:16 sleep tips, 10s (30 Oct) | 35.6 | ~4m55s, incl. 3 nudges | 1080x1920, native vertical layout, audio present. Ran 15.1s. |
Exports took about 30 seconds to 1m40s to render. In total the 12 jobs in rounds 1–3 used about 313 credits.
Round 4 (30 Oct). Every job asked for narration, music and captions (except J8, which asked for no captions) and a set length. J1–J6 ran at the same time and used 311.2 credits between them; J7 and J8 ran alone.
| Job | Credits | Generation time | Result |
|---|---|---|---|
| J1 PowerPoint: coffee deck, 20s | batch | ~10 min, incl. failure and retry | Failed once, 2 nudges (one with leaked error text). 20.0s, no audio, no captions. |
| J2 PowerPoint: EV deck, 20s | batch | ~11 min, incl. failure and retry | Failed once. 20.0s, no audio, no captions. |
| J3 Document (.md): remote work, 20s | batch | ~10 min, incl. failure and retry | Failed once. 20.0s, no audio, no captions. |
| J4 Document (.docx): plastic-free kitchen, 20s | batch | ~10 min, incl. failure and retry | Failed once. 20.0s, no audio, no captions. |
| J5 URL: aividlab.co, 20s | batch | ~7 min | Audio and small captions present. Ran 24.0s. |
| J6 URL: Wikipedia “Aurora”, 20s | batch | ~8 min, plus ~1.5 min page scrape | Audio and captions present. Ran 32.8s. |
| J7 Product ad from an image, 15s | 42.1 | ~6m25s, incl. a stall | Audio and captions present. Ran 19.8s. |
| J8 Prompt: why the sky is blue, exactly 20s, no captions | 61.2 | ~7m30s | Audio present and correctly no captions. Ran 27.8s. |
Round 4 exports took about 30 seconds to 1m20s to render. Across rounds 1–4 the 20 jobs used about 727 credits; the two round-5 jobs used 53.4 more.
Output & Export
- Resolution: 720p or 1080p only. There’s no 4K option in the export dialog, and the pricing page lists 1080p downloads for Pro and Ultra.
- Formats: MP4, WebM or MOV, with Draft, Standard or High quality at 24, 30 or 60fps.
- Watermark: none on any of our paid-plan exports. The pricing page says the free plan exports 720p with a watermark.
- Aspect ratio: set in the chat brief, not at export. 16:9 is the default, and 9:16 worked when we asked for it. We didn’t test other ratios.
- Audio: every PowerPoint and document job with a voiceover (2 on 28 Oct, 6 on 30 Oct, including two run one at a time) exported with no audio track, and the 30 Oct ones had no captions either. Every web-page, image and prompt job with narration had a working audio track. Play each export with sound before you publish it.
Credits & Pricing (checked 30 Oct 2026)
| Plan | Monthly | Annual | Credits | Downloads |
|---|---|---|---|---|
| Free | $0 | $0 | Credit box shows “0 Credits”; feature list says “10 free credits” | 720p, with watermark |
| Pro | $29/mo | $21/mo ($252 billed annually) | 500/month | 1080p, no watermark, priority support |
| Ultra | $69/mo | $49/mo ($588 billed annually) | 1,400/month | 1080p, no watermark, priority support |


Prices from cutbay.ai/pricing, both billing toggles checked 30 Oct 2026. Pro and Ultra also offer other credit tiers in a dropdown, which we didn’t open. Open the page for live numbers.
Cost per Video (our estimate)
We base this on the two round-4 jobs that ran alone and didn’t fail, so their credit cost is clean: a 19.8-second product ad used 42.1 credits and a 27.8-second narrated explainer used 61.2 credits (the two columns below). That’s about 2.1–2.2 credits per second of finished video.
| Plan | Per credit | 19.8s ad | 27.8s explainer | Videos/mo |
|---|---|---|---|---|
| Pro, monthly ($29 / 500) | $0.058 | ~$2.44 | ~$3.55 | ~8–11 |
| Pro, annual ($21 / 500) | $0.042 | ~$1.77 | ~$2.57 | |
| Ultra, monthly ($69 / 1,400) | ~$0.049 | ~$2.07 | ~$3.02 | ~22–33 |
| Ultra, annual ($49 / 1,400) | $0.035 | ~$1.47 | ~$2.14 |
CutBay’s pricing page estimates about 25 videos a month on Pro and about 70 on Ultra. For 20- to 30-second narrated videos like ours, expect roughly a third to a half of that. A failed run that needs a retry is charged again, so a job that fails once costs about twice as much.
Our arithmetic from 30 Oct 2026 prices and our own credit logs; credits spent on the agent’s questions are included. Shorter clips, templates and the Talking Head Editor cost less in our tests.
Pros
- Web-page, prompt and image jobs turned into tidy narrated videos with accurate on-screen text
- 1080p exports in MP4, WebM or MOV, with no watermark on paid plans (22 of 22 jobs)
- Many ways in: prompt, URL, image, PowerPoint, document, template remix
- Native 9:16 layouts when you ask for vertical video
- Built-in voices, music and captions in one pass
- Talking Head Editor subtitles uploaded footage quickly, with no visible credit charge in our test
Cons
- Every PowerPoint and document job with a voiceover (8 of 8) exported with no audio and, on 30 Oct, no captions
- Every deck and document job on 30 Oct (6 of 6, including two run alone) failed at “Designing scenes”, and retries are charged again
- The agent still stalls and needs nudges, sometimes showing internal error text
- Loose on instructions: lengths often overshot (by up to 68%), and “no music” or “no subtitles” was often ignored
- Real cost per 20–30s video is about $1.50–$3.60, so you get far fewer videos than the plan estimate
- No stock avatars or lip-sync, despite the “Talking Head” name
- Not a generative video model: motion is mostly stock clips, stills and pans
- Max 1080p, no 4K
Who It’s For
- Good fit: marketers, educators and small teams who want a short narrated explainer or social clip, including 9:16, from a web page, an image or a prompt, without editing.
- Not yet: turning PowerPoint decks or documents into narrated videos. Every one we ran exported silent. If you try it, check the sound on the export.
- Good fit: creators who record themselves and want quick subtitles and music on that footage.
- Look elsewhere: cinematic or camera-controlled AI video (Higgsfield, Kling AI), avatar presenters (Synthesia, Vidnoz AI), or blog-to-video with more manual control (Pictory AI).
Before You Buy
- Try the free plan on your real material first (720p with a watermark). If it’s a deck or document, check the export has sound.
- Don’t count on running jobs one at a time to fix deck or document jobs: our solo re-runs failed and exported silent too.
- Put the aspect ratio, exact length, “no subtitles” and “no music” in your first prompt, and check the brief before you confirm.
- If the agent goes quiet or shows error text, reply “please continue”. That fixed every stall we hit. If a job shows “Designing scenes” failed, a chat message asking for a retry fixed it, but you pay again.
- Play the export with sound before you publish it.
Verdict
CutBay can turn a web page, an image or a short prompt into a tidy 1080p explainer with voiceover, music and captions, and paid plans carry no watermark. But the deck and document workflow it leads with isn’t dependable yet: all eight PowerPoint and document jobs we ran with a voiceover exported silent, including two run one at a time, half of our round-4 jobs failed and were charged again on retry, and lengths and caption instructions were loosely followed. At about $1.50–$3.60 per 20–30-second video (30 Oct 2026 prices), it’s worth a free-plan trial for URL or prompt explainers, but we’d wait for fixes before paying for docs-to-video work.
Frequently Asked Questions
Try CutBay
Agent-style AI video creator for turning web pages, prompts and images into short 1080p videos.
Visit CutBay →See affiliate disclosure
Update Notes
November 6, 2026: First published. 22 hands-on jobs over five sessions on 28 and 30 Oct 2026, with both pricing toggles checked 30 Oct 2026.