Our own test

Same character, same line: AvatarSay vs KreatorFlow vs Dreamina

Run by the AvatarSay team on 7 November 2026 one run per tool competitor free credits, no top-ups

AvatarSay could not make this clip on its own free tier. Its free first clip now runs up to 5 s and 75 characters (raised from 3 s on 7 November 2026, the day of the test). The test line is 83 characters and 5.8 s of speech, so it still doesn’t fit, and the 30 sign-up credits are below the 72 it needs. On AvatarSay this clip costs 72 credits: $1.43 at Starter and $1.06 at Studio. KreatorFlow and Dreamina made it on sign-up and free credits for $0.

AvatarSay was also the slowest. Its 201 s run went through the API directly, which skips the product’s job queue; the AvatarSay site says renders usually take 5–15 minutes. Our character also raised both hands into frame while talking.

Three frames of the same cartoon grandpa inventor at 2.5 seconds, left to right: AvatarSay, KreatorFlow and Dreamina. — Frame at 2.5 s. Left to right: AvatarSay, KreatorFlow (Seedance 2 Fast), Dreamina (Avatar Turbo). Stills only; we don’t republish the other tools’ videos.

A real free-clip test · November 12, 2026

From a description to a talking fox

One operator test using a Standard image and the free first clip. Results and waiting times vary; audio quality and lip sync were not independently evaluated.

The selected Standard image of a fox spaceship captain

Selected Standard image

Unedited output · 2.93 seconds · Open MP4

Character prompt: A nervous fox spaceship captain, facing forward, with a visible mouth and a simple background
Style and line: 3D Cartoon, Standard · “Welcome aboard. Try not to press that.” · Friendly voice
Cost and timing: The image pair cost 4 credits; the first clip used the free offer. This single video task took 3 minutes 32 seconds from server creation to success, not a typical or guaranteed wait. One video attempt, no retry.

New accounts receive 30 credits, which can be used to create images. Separately, accounts that have never purchased credits may receive one free first clip of up to 5 seconds and 75 characters, subject to the limited launch allocation and daily and IP limits. Availability is shown in the studio; longer spoken audio may not qualify.

AvatarSay creates new fictional characters from text. It cannot animate an uploaded photo or an existing character image.

The fox above is a separate free-clip product test. The comparison on this page uses a different image and an 83-character line; AvatarSay’s comparison clip required paid credits and was run directly through the API, without the product queue. The two tests are not directly comparable.

Method

How we tested

Character: AvatarSay’s 3D Cartoon sample 3d-2, the prompt “a grandpa inventor with goggles and a bushy mustache”. The same file went to every tool. See the source image.
Line: “Hi, I'm Grandpa Gizmo. Today we're fixing a squeaky door, one tiny screw at a time.”
Voice: each tool’s own. Casual Guy on AvatarSay; KreatorFlow generated the voice from the model with no voice picker; “Bill” on Dreamina.
Wall-clock: from pressing generate (or starting the script) to a finished file.
Transcript: fal Whisper on each output’s audio track.
One run per tool. Competitor accounts used free credits: no trials, new terms or payment details. All outputs are unedited.
Spend: about $0.45 in model costs on our side; $0 on the competitors.

Results

What each run produced

Same-prompt test results for AvatarSay, KreatorFlow, Dreamina and Hedra
Measure	AvatarSay	KreatorFlow	Dreamina	Hedra
Ran?	Yes	Yes, but not with Lip Sync	Yes	Blocked
Tool / model	MiniMax speech-2.8-turbo + Kling AI Avatar v2 Standard	Seedance 2 Fast, image to video, audio on	Avatar Turbo (“OmniHuman 1.5”), voice “Bill”	Hedra Avatar (not run)
Length	7.2 s (5.8 s speech; Kling pads the end)	5.1 s	6.6 s (UI said 7 s)	—
Resolution / fps	960×960, 30 fps	960×960, 24 fps	960×960, 60 fps	—
Says the full line?	Yes	Yes	Yes	—
Wall-clock	201 s (direct API run, no job queue). The site says renders usually take 5–15 minutes.	~115 s	~124 s	—
Credits / cost	72 credits = $1.43 at Starter, $1.06 at Studio. Our fal cost ≈ $0.42	12 of the 25 sign-up credits. $0	42 shown; 96 → 47 balance (49 deducted). $0	—
Possible on the free tier?	No. The line (83 characters, 5.8 s) is over the free clip's 5 s / 75-character limit, and 30 sign-up credits are below 72.	Yes, via the one free first video (Lip Sync is paid)	Yes	No (0 credits)
Watermark	None	None seen (four corners checked)	Yes: “AI” badge top-left + “Dreamina AI” logo bottom-right	—

One by one

What each tool did

AvatarSay

Our AvatarSay output, re-encoded to 720×720.

AvatarSay ran the same models its product uses: MiniMax speech-2.8-turbo and Kling AI Avatar v2 Standard. The output was 7.2 s at 960×960 and 30 fps, with 5.8 s of speech; Kling pads the end. The transcript had the full line. Wall-clock was 201 s on the API direct path.

The price is 72 credits: $1.43 at Starter, $1.06 at Studio; our fal cost was about $0.42. There is no watermark. It keeps the source framing, and the character raises both hands into frame while talking.

KreatorFlow

Frame from KreatorFlow's output: the cartoon grandpa inventor speaking with an open mouth, no hands in frame.

Lip Sync was blocked on the free account: “Upgrade to get access to Lip Sync”. We used the free first video instead: Seedance 2 Fast, image to video with “Audio on”, 5 s, 720p, 1:1. That is KreatorFlow’s normal UI, but it is not its Lip Sync, and the voice is not selectable.

The output was 5.1 s at 960×960 and 24 fps, with the full line, in about 115 s, for 12 of the 25 sign-up credits ($0). No watermark was visible in the four corners. Its real-face notice fired on our 3D cartoon (“This photo looks like a real person...”), but generation still completed. The frames show open-mouth shapes, no hands, and the source framing.

Dreamina

Frame from Dreamina's output: the cartoon grandpa inventor speaking, framed tighter with the top of his head cropped.

Dreamina used existing free credits: AI Avatar, Avatar Turbo (“OmniHuman 1.5”), voice “Bill”. The output was 6.6 s (the UI said 7 s) at 960×960 and 60 fps, with the full line, in about 124 s. The UI showed 42 credits; the balance went from 96 to 47, so 49 were deducted. $0.

It carries an “AI” badge top-left and a “Dreamina AI” logo bottom-right. It reframed tighter, with a larger face, the top of the head cropped closer and more facial expression (squints and closed eyes near the end). The upgrade offer and role survey were dismissed without accepting anything.

Hedra: not tested, and why

Hedra was not tested. The account showed the Free plan with 0 credits, and its pricing page lists no free plan with credits. After the image and voice upload, a biometric consent dialog appeared: facial-geometry scans, voice data processing, a third-party disclosure and a confirmation that the upload contains no human other than yourself. We clicked “I do not agree”.

Even with consent, 0 credits would block generation. To test it we would have to accept that consent and buy credits (Basic is $20/month) or use the prepaid API. There is no Hedra output, so we say nothing about its quality.

Choosing

What this means if you’re choosing

If $0 and a short clip are the whole requirement: KreatorFlow’s free first video (5.1 s) and Dreamina’s free credits (6.6 s) were enough, and both had the full line. For KreatorFlow, remember it was Seedance video, not Lip Sync, with no voice picker; its real-face notice fired on a cartoon.
If a watermark decides it: AvatarSay had none, and none was seen on KreatorFlow’s output. Dreamina carries the “AI” badge and the “Dreamina AI” logo.
If framing must match the source: AvatarSay and KreatorFlow kept it; Dreamina reframed tighter.
If you want AvatarSay: this line is a paid, per-clip purchase of 72 credits: $1.43 at Starter and $1.06 at Studio. It is not possible on the free tier. You get no watermark and the source framing, at the cost of time (201 s via the API; the site says “usually 5–15 minutes”) and the hands in frame.

Limits

What this does and doesn’t show

One run per tool, so output varies from run to run.
One character and one line; this says nothing about other images, voices or scripts.
Voices differ by design. Audio quality and lip-sync accuracy were not judged by a listener.
Visual notes come from frame grabs.
KreatorFlow’s result is a different feature from its paid Lip Sync.
AvatarSay’s 201 s is the API path, not the queued product path.
We did not score any output.

More on these tools

KreatorFlow alternative Dreamina alternative Talking cartoons

HeyGen, D-ID, Hedra, KreatorFlow, Dreamina / CapCut and MonkeyTalker are trademarks of their owners. AvatarSay is not affiliated with them.