Gen keeps some resemblance to the original speaker on the noisy source sample, but it only partially preserves identity and still sounds robotic.
What was measured
Voice Match Accuracy
Whether the generated voice actually sounds like the original speaker, including tone, pitch, rhythm, and identity.
decisive for this rankingtransformation
This ranking is fundamentally about whether the generated speech still sounds like the target speaker, so identity match is core. (3 of 3 judges)
What was given, what came back
Test input: Low-Quality Voice Sample · mixed · group: voice-cloning
Input — what we sent
Input, verbatim
Removing objects from videos used to take hours of manual editing. Now AI tools claim to do it in minutes. So we tested five AI video object removers to find the most reliable one. We used the same three inputs across all the tools for a fair comparison. ABC Labs showed unstable tracking and heavy distortion. Media.io offered fast processing but unusable outputs. PhotoRoom mostly relied on blur masking instead of real reconstruction. Runway delivered the cleanest removals with the most stable tracking and realistic scene reconstruction. Here's exactly how we tested it.
0:00 / 0:00
Loading audio...
0:00 / 0:00
Loading audio...
Low-Quality Voice Sample
A noisy voice recording with background noise, room ambience, and minor disturbances, used to test whether voice-cloning tools can preserve speaker identity when the source audio is imperfect.
Why this input is hard
- · Cloning accuracy from degraded audio
- · Noise and ambience robustness
- · Speaker identity preservation under poor recording conditions
- · Distinguishing enhancement from true cloning
Output — unretouched
0:00 / 0:00
Loading audio...
Also checked on this input — same tool, 9 other criteria
Long-Form Consistency✓ WorkedGen stays consistent through longer passages on the noisy sample, with no major voice breaks, glitches, or abrupt tonal shifts.Long-Form Consistency✓ WorkedHD stays consistent through long-form narration on the noisy sample, with no abrupt changes in voice quality or pronunciation.Long-Form Consistency✓ WorkedGen+ remains stable through long-form generation on the noisy sample, with no noticeable interruptions or voice degradation.Naturalness & Human Quality✓ WorkedHD is the most human-like output on the noisy sample, with better emotional tone, speech rhythm, and vocal realism than Gen or Gen+.Naturalness & Human Quality◐ MixedGen+ is more natural than Gen, but the gender inconsistency keeps the overall cloning quality from feeling convincing.Naturalness & Human Quality⚠ StruggledGen sounds noticeably robotic and lacks emotional depth and natural speech rhythm on the noisy sample.Pronunciation Accuracy✓ WorkedGen keeps the standard script intelligible on the noisy sample; the robotic tone is a delivery issue, not a misread-word problem.Pronunciation Accuracy✓ WorkedHD delivers the best pronunciation result on the noisy sample, with no misread or garbled words noted during the listening pass.Pronunciation Accuracy✓ WorkedGen+ keeps the script intelligible on the noisy sample despite the gender-shift issue, and the report notes no dedicated pronunciation stress test for this scenario.
Provenance
- Observation
- 86b07cc2-a3a8-4612-85ce-cef2a6853f8a
- Evidence run
- 46222c41-0046-41cc-bfaa-5f7ba6aa4933
- Study
- Clone Your Voice and Generate Voiceover from Text
- Research task
- 86ba42bx1
- Tested at
- not recorded
- Source
- first-party
- Evidence state
- verified
- Proof shown
- input + output shown
- Cost / latency
- not captured
- Repeat run
- not captured
- Tester
- not captured
The last three rows are honest blanks, not placeholders — our capture has no field for them yet.
Query this
get_evidence({
tool: "topmediai-voice-cloning",
scenario: "voice-cloning"
})MCP · mcp.aidemos.com/api/mcp
Free with attribution.
Same input, same check — 9 other tools
measured on Voice Match Accuracy
AICloneVoiceFree.com✓ WorkedVoice cloning stays strong even from a noisy source, with the report estimating about 95% similarity to the original speaker and preservation of most vocal characteristics.ElevenLabs◐ Mixed50Clones only about half of the source-speaker identity in the noisy sample: the report rates the match at approximately 50% and says the output sounds heavily polished, which lowers resemblance instead of faithfully reproducing the original voice.Fish Audio✓ WorkedPreserves speaker identity well even from a noisy source: both low-quality English outputs were described as strong, close matches to the original voice.Heygen◐ MixedA mid-tier low-quality clone reached roughly 70–80% similarity to the original voice, making it acceptable for short-form use but still not fully faithful.Inworld◐ MixedOn the ~53-second low-quality clone, identity match was strongest at the start and then faded gradually as the clip progressed, so the voice was only partially consistent rather than exact.MiniMax◐ MixedWith a noisy source recording, the clone only partially preserved speaker identity: the generated voice was noticeably softer than the original and the report rates the match as Fair (~35–45%).Speechify✗ FailedOn the noisy source, Speechify produced a clone that diverged strongly from the speaker and even sounded female despite a male input.Uberduck✗ FailedThe noisy-source clone barely resembled the original speaker and was described as the weakest voice-match result in the round.VocalAI⚠ StruggledThe clone preserved only about 10–15% of the original speaker's identity, and the output sounded heavily polished and processed rather than speaker-faithful.
This evidence is published in
Real inputs and real outputs, no retouching · every cell queryable via API & MCP · aidemos.com